A POSITRON DETECTOR FOR PRECISION
BETA DECAY EXPERIMENTS FROM A
MAGNETO-OPTIC TRAP
Dan G. Melconian B.Sc., MCMaster University, 1995
A thesis submitted in partial fulfillment of the requirements for the degree of
MASTER OF SCIENCE in the
Department of P hysics
@ Dan G. Melconian 2000 SIMON FRASER UNIVERSITY
July 2000
Copyrights are not reserved. Permission is hereby granted to
reproduce this work in whole or in part.
sitions and Acquisitions et B' iogiaphic Senrices senrices bibliographiques "ds"
The author bas granteci a non- L'auteur a accordé une licence non exchuive licence aüowing the exclusive permettant à la National Li'brary of Canada to Bibliothèque nationale du Canada de reproduce, loan, distri'bute or seii reproduire, prêter, distribuer ou copies of this thesis in microform, vendre des copies de cette thèse sous paper or electronic formats. la fonne de microfichelfilm, de
reproduction sur papier ou sur format électronique.
The author retains ownership of the L'auteur comme la proprieté du copyright in this thesis. Neitha the &oit d'auteur qui protège cette thèse. thesis nor substantial extracts fiom it Ni la thèse ni des extraits substantiels may be printed or othenivise de celle-ci ne doivent être imprimés reproduced without the author's ou autrement reproduits sans son permission. autorisation.
Abstract
The TRINAT collaboration has rnagneto-optically trapped 38mK from TRIU~IF'S ra-
dioactive beam facility with the goal of measuring to better than 1% the b - t~ cor-
relation parameter, a, of the superallowed P decay. This measurement will test the
Standard Mode1 prediction of a = +l, where a deviation from unity wvould be a clear
indication of scalar contributions to the weak interaction.
The subject of this thesis is the design, optimization and characterization of a
plastic scintillator used in coincidence with a double-sided Si-strip detector to observe
the E and AE of the emitted positron. The timing response of the scintillator relative
to a rnicrechannel plate has a width of o = 1 ns, which provides a good time-of-flight
measurement used to determine the recoil momentum, allowing that of the neutrino
to be deduced.
The scintillator was designed with the aid of Monte Carlo simulations and opti-
mized for the O - 5 MeV region oE interest. The energy has been calibrated to within
f 5 keV using the Compton edges of y sources as well as by the on-line B spectrum.
The gain of the scintillator is kept constant over the course of the experiment to within
0.15% by a stabilization system.
The silicon strip detector has been calibrated using photon sources (30-80 keV)
and the calibration was extended using the on-line data by requiring energy agree-
ment between the 2 and jj strips. Using these calibrations, a detailed analysis scheme
has been developed that accounts for charge sharing between strips and multiple hits.
The strip detector is an important component of the telescope because it provides an
effective tag for B events which are needed to discriminate against the 38K Y back-
ground in the scintiiiator. A hardware coincidence behveen the E and A E detectors
reduces this background by a factor of 35, and the analysis scheme reduces it another
order of magnitude with minimal loss of good P events.
The P telescope's energy spectrum from the April-May 1999 experirnental run is well reproduced by detailed simulations above 2.5 MeV, the lowest P energy ex-
pected to be used in the analysis of a. The simulations are not quite as accurate in
reproducing the background and the Compton summing of the annihilation radiation,
limiting our understanding of the lower-energy part of the 0 spectrum. A coincidence
condition with the recoil detector, however, virtually eliminates these backgrounds,
providing a clean measurement for the correlation experiment.
In memory of Otto Hausser
There are grounds for cautious optimism that Ive may now be near the
end of the search for the ultimate laws of nature.
Stephen Hawking
A Brief Histoy of Time, 1988
Acknowledgement s
1 would like to thank first and foremost the late Prof. Otto Hausser for providing
me with the wonderful opportunity of working with TRINAT as his student as well
as for instilling in me a part of his deep love for physics. 1 only wish ive could have
worked together longer because 1 know there was still much more that 1 could have
learned. Thanks to the intervention of Peter Jackson who took over the responsibility
of supervising me, 1 was able to continue my studies. In addition to helping me with
analysis and offering ideas on a daily basis, his comments and suggestions on countless drafts of this thesis have been used throughout and have been an immense belp to
me. John Behr also deserves much credit, for he was the one who first helped me
fumble my way through TRIWAT'S complicated hardware, and was instrumental in
the testing and characterization of the scintillator. He has always played an active
d e in helping me with analysis, performing complementary calculations and taking
the time to talk about physics.
1 would also like to thank my supervisory cornmittee, Byron Jennings, Howard
Trottier and Mike Vetterli for taking an interest in my work and for the help they
have given me.
My fellow TIUNAT grad students have made the day- tday routine mucb more
pleasant, especially on those the long, stressful days before and during runs. 1 owe
a lot to Alexandre Gorelov for his work that is used in this thesis which includes
(but is not limited to): his recoil tracking code for GEANT; the new DSSSD mount;
Figures 3.6, 4.6, (the cube of) 3.4 and B.l; aiso, he has continually helped me with
setting up and optimizing the hardware. 1 have also learned many cornputer and 'shop' skiils from him, but perhaps most importantly, 1 would like to thank him for
always being up for a coffee/smoke break; 1 have valued our conversations together
vii
whether they were about physics, current news, or yes, even the war in the Balkans!
Mike Ttinczek has b e n a great help tù my mental health by always staying positive
and taking the time to chat with me about physics as well as life in general. He
also tried tu foster my physical activity through voiieyball, basketbal1 and succer, but unfortunately, he's only human. As a result of his use of GEANT, a number of bugs
were discovered in the code and fixed, and 1 thank him for useful comments especially
with regard to appendix B-2. Good luck with writing up, guys!
1 would like to thank the following members of the group for being additional
sources of information and guidance. Ulrich Giesen, who set up the hardware for
the DSSSD and is responsible for the basis of TRINAT'S hardware as a whole, always
left his door open and my only regret is 1 didn't use it often enough while he was in
Vancouver. I owe a lot of the clarity in Chapter 2 to Pierre Dubé who was alwvays
interested in my work even though it wasn't atomic physics. 1 have had many useful
discussions with Parker Alford who has also helped editingdrafts of this thesis. 1 \vould
also like to thank Trevor Stocki for Looking the thesis over, especially appendk B.L. 1 clid not get a chance to spend as much time with the Jens' that passed through
the group (Dilling and Schmid), but am happy to have had the pleasure of wvorking
with them. I: would like to extend a special thanks to Jens Dilling for making a special
effort to make me feel welcome at TRIUMF when 1 first joined the group.
Many thanks needs to go to Steve Chan and the rest of the guys in the scintillator
shop. If not for their patient natures, flexibility with rushed orders, and extended
experience with scintiiiators, the one used in this thesis surely tvould not have been made as quickly or with as good a quality.
1 would also like to acknowledge various help and useful cliscussions from John
DYAuria, Guy Savard, Tom Davinson, Ted Clifford, Pierre ,haudruz, Renée Poutis-
sou, Peter Machule and Jimmy Chow. The stabilization of the scbtillator would not
have been possible without the generosity of York Holler who has lent TRINAT two of
his units. Joe Chuma has written an excellent program in physica, which has been
used throughout this thesis. Also, thanks to Anna Gelbart for providing me with the
basis for Figure 3.5.
The graduate secretary, Candida Mazza, has shielded me Erom much of the pa-
p e m r k and ha kept track of important dates for me; thank you for Lettiug me
concentrate on physics rather than worrying about bureaurocratic details. Finaity, 1 would like to thank Jan Blanchard, my family and my friends for their
continued support and for always believing in me. Jan, you have been my best üiend since the McMaster days, and always having you around through the good times and the bad has made living here in Vancouver infinitely easier. Thank you for taking a genuine interest in my work and for keeping my head on my shoulden these last few
years! Mom, babi, L m n and Sonja: there is no way 1 would have been here if not for your love, faith and support - 1 hope I've made you proud!
Table of Contents
Approvai
Abstract
Acknowledgements
Table of Contents
List of Tables
List of Figures
. . 11
iii
vii
X
xiii
xiv
Chapter 1 Introduction 1
Chapter 2 /3 Decay and Fundamental Symrnetries 4
2.1 The Fermi Mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 A Generalized Interaction . . . . . . . . . . . . . . . . . . . . . . . . . 10
Chapter 3 The /3 - v Correlation Experiment 14
3.1 Using 38mK to rneasure a . . . . . . . . . . . . . . . . . . . . . . . . . . 14 3.2 Tkapping Techniques . . . . . . . . . . . . . . . . . . . . . . . . - . . . 19
3.2.1 The Neutrd Atom Trap . - . . . . . . . . . . . . . . . . . . . . 19 3.2.2 TRIUMF'S Radioactive Potassium Source . . . . . . . . . . . . . 23
3.2.3 TRINAT'S Double MOT System . . . . . . . . . . . . . . . . . . 26
3.3 Nuclear Detection System . . . . . . . . . . . . . . . . . . . . . . . . . 28
Chapter 4 The Positron Detector 3 1 . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Design and Construction 31
. . . . . . . . . . . . . . . . . . . . . . . 4.1.1 Design Considerations 31 . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2 GEANT Simulations 32
. . . . . . . . . . . . . . . . . 4.1.3 The Final Design of the Telescope 37
. . . . . . . . . . . . . . . . . 4.2 The Double Sided Silicon Strip Detector 41 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 TheDevice 41
. . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2 Energy Calibration 43 . . . . . . . . . . . . . . . . . . . . . . . 4.2.3 Extended Calibrations 46
. . . . . . . . . . . . . . . . 4.2.4 Characterization of the Resolution 51 . . . . . . . . . . . . . 4.2.5 Position Decoding and Analysis Scheme 56
. . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.6 38mK Results 60 . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 The Plastic Scintillator 62
. . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Optimization 63 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Timing 64
. . . . . . . . . . . . . . . . 4.3.3 Energy Calibration and Resolution 66 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.4 Stabilization 74
. . . . . . . . . . . . 4.3.5 38mK Results and an Extended Calibration 78 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 The Ptelescope 86
. . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Total p Energy 88 . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2 Backscattering Losses 93 . . . . . . . . . . . . . . . . . . . . . . . 4.4.3 Uniformity of Response 94
. . . . . . . . . . . . . . . . . . . . 4.4.4 The Fierz Interference Term 95 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 @-Ar Coincidences 96
. . . . . . . . . . . . . . . . . . . . . . . . . . 4.5.1 Scattering Efïects 97 . . . . . . . . . . . . . . . . . . . . 4.5.2 Recoil Coincident ,f3 Spectra 102
Chapter 5 Conclusions 106
Appendix A Response function of the scintillator 109 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.1 Saturation Mects 110
TABLE OF CONTENTS xii
A.2 Bremsstrahlung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 A.3 Annihilation Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Appendix B GEANT and Future Work 117 B.1 Future geometries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 B.2 Massive neutrinos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Appendix C Electronics of the bTelescope 122
Bibliography
List of Tables
Possible forms for an interaction consistent with Lorentz invariance . . . 5
. . . . . . . . . . . . . . . . . . . . . ISAC radioactive beam intensities 26
Low-energy photon sources used to provide an initial calibration of the . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . strip detector 44
Fit parameters of the DSSSD stries off-line source calibration . . + . . 47 DSSSD calibration fits of E4i = (E, ) = crri x EXi + Px. . . . . . . . . . 50
. . . . . . . . . DSSSD calibration fits of E;i = (Ex) = ayi x Eyi + Py. 52
. . . . . . . . . . . . . . . . Final midths of the DSSSD energy readings 55
. . . . . . . . . . . . . . . . . y calibration of scintillator for April 1999 71
Map of the X' dependence on the low-energy cut-off of the fitting region 84
. . . . . . . . . . . . . . . . GEANT calculations of ,d scattering effects 100
List of Figures
2.1 Fermi's mode1 for B+ decay . . . . . . . . . . . . . . . . . . . . . . . . 7 2.2 General Feynman diagrams for a Of -4 O+ B+ decay . . . . . . . . . . 11
3.1 Nuclear energy levels for 38mK . . . . . . . . . . . . . . . . . . . . . . . 15
3.2 Recoil time-of-flight vs . B energy in a back-to-back geometl . . . . . . 16
3.3 Atomic energy levels of 38mK in a l-D MOT . . . . . . . . . . . . . . . 22
3.4 Schematic diagram of a vapour-ce11 MOT and neutralizer . . . . . . . . 24
3.5 The Is~c radioactive beam facility . . . . . . . . . . . . . . . . . . . . 25
3.6 Schematic diagram of TRTNAT'S ,4? - v correlation experirnent . . . . . . 27
3.7 Schematic diagram of an MCP detector . . . . . . . . . . . . . . . . . . 29
4.1 GEANT response functions of a plastic scintillator . . . . . . . . . . . . 33 4.2 MC of events lost due to multiple scattering in the DSSSD . . . . . . . 34
4.3 MC simulations depicting where positrons annihilate in the scintillator . 35 4.4 MC design simulations of multiple scattering effects due to the /3 window 37 4.5 Schematic diagram of the P-telescope assembly . . . . . . . . . . . . . . 38
4.6 Schematic diagram of the DSSSD mounting . . . . . . . . . . . . . . - 39
4.7 Geometry of TRINAT'S detection chamber input into GEANT . . . . . . 40
4.8 Schematic diagram of a p+n double-sided silicon strip detector . . . . . 42
4.9 Sample fits to the DSSSD energy spectrum of *"Am . . . . . . . . . . 45
4.10 Energy calibration of the DSSSD strips x2 and y2 using 133Ba and '"-Am 46
4.11 Comparison of Ex and Ey using the off-line calibrations . . . . . . . . . 50
4.12 Comparison of Ey, and Ey, to the corrected E: . . . . . . . . . . . . . 52
4.13 Plot of E:,, vs . (E;) - E;,, . . . . . . . . . . . . . . . . . . . . . . . . 03
4.14 Resolution hnction for strips xl. x12, y1 and y12 . . . . . . . . . . . . . 54
4.15 Block diagram of the DSSSD analysis scheme . . . . . . . . . . . . . . 56
4.16 PositionofDSSSDB hitsinz . . . . . . . . . . . . . . . . . . . . . . . 59
4.17 Average position of inter-strip DSSSD events . . . . . . . . . . . . . . . 60
4.18 DSSSD energy spectrum of the on-line 38mK . . . . . . . . . . . . . . . 61
4.19 207Bi spectra using different scintiilator wrapping schemes . . . . . . . . 64
4.20 Scintillator-MCP timing . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.21 Kinematics of the Compton effect and the scattered electron's spectrum 67
4.22 Fits of a 88Y Compton spectrum to a GEANT simulation . . . . . . . . 70
4.23 Calibration of May 2nd, 1999 using the Compton edges of y sources . . 73
4.24 Non-linear fit to the scintillator's calibration y sources . . . . . . . . . 73
4.25 Resalution of the scintitlator as determined by fits to Compton edges . 75
4.26 Stabilization test of the scintillator's gain . . . . . . . . . . . . . . . . . 76
4.27 Long term test of the scintillator stabilization system . . . . . . . . . . 77
4.28 Scintillator Kurie plots for different AEDSSSD conditions . . . . . . . . . 80
4.29 Fit of the 38mK spectrum to a MC simulation . . . . . . . . . . . . . . 82
4.30 Fit of the 38mK spectrum above 2.2 MeV to a MC simulation . . . . . . 87
4.31 The telescope's B spectra for different AEDsssD conditions . . . . . . . 89
4.32 Fit of the 38mK Tp spectrum to a MC simulation . . . . . . . . . . . . . 91
4.33 Fit of the 38mK Tp spectrum above 2.3 MeV to a MC simulation . . . . 92
4.34 Uniformity of response of the scintillator . . . . . . . . . . . . . . . . . 95
4.35 MC simulation of the recoii TOF versus B energy for ArfL recoils . . . 98
4-36 Scatter plot of recoil TOF versus Tp from the on-line 38mK . . . . . . - 101
4.37 TOF projections for the ArfL and comparison to GEANT . . . . . . . -101
. . . . . . . 4.38 The 38mK /3 spectra gated on Ap?+1-+2 recoil coincidences 103
4.39 The 38mK P spectrum gated on .Ar f 3 p H recoil coincidence . . . . . . . . 104
A l Total cross-sections for the energy loss of positrons in plastic . . . . . . 111
A.2 MC simulation of saturation effects in plastic . . . . . . . . . . . . . . . 112
A.3 Radiative energy loses in plastic ancl silicon . . . . . . . . . . . . . . . 114
B.1 Schematic diagram of TRINAT'S new electmstatic hoop design . . . . . 118
B.2 Kinematics of 38mK decay with massive neutrinos . . . . . . . . . . . . 120
LIST OF FIGURES xvi
C.1 Electronics diagram for the P-telescope . . , . . . . . . . . . . . . . . . 123
Introduction
As Hawking and other physicists believe (albeit with reservation), we appear to be
very close to finally understanding the fundamental laws governing the universe. Im-
proved technologies have allowed measurements that are testing to greater and greater
precision the current theory of particle physics: the Standard Model. This mode1 is
comprised of quantum electrodynamics (describes electromagnetic processes), the elec-
troweak theory of Glashow, Weinberg and Salam (unifies weak and electrornagnetic
forces) and quantum chromodynarnics (governs strong interactions). The Standard
Model cannot be the 'final theory,' howvever, because i t cloes not unify the electroweak
and strong interactions. A 'Grand Unified Theory' is one in which al1 three interac-
tions are seen to be low-energy manifestations of a single force. The next (and perhaps
final?) step would be to include gravity so that al1 four forces would be described by
one ultimate theory.
Hawking retains "optimi~m'~ that a unified theory wilI one day soon be discovered.
Perhaps it will be; or, Like at the end of the 1800'~~ perhaps it will be that ive will
find evidence of new physics, this time fiom results outside the Standard Model. It
is my "optimistic" hope that this wilI be the case so that the physics community \vil1
have many new questions to answer for a long time to come.
The Standard Model has to date stubbornly resisted any attempts at proving it
wrong; it remaïns one of the most thoroughly tested models in science, and has not
failed yet. It is important to continue searching For physics beyond the Standard
Model because any deviation wodd be an important guide for, or test of, a unified
t heory.
Precision /3 decay experiments can be a sensitive test of the Standard Model be-
CHAPTER 1 INTRODUCTION 2
cause they are inherently weak processes. The novel technology of neutrai atom traps
has opened the door to a new generation of /3 decay experirnents. TRINAT has utilized
this technology and is currentLy rneasuring the B - v correlation parameter, a, in the
0'- O+ ,û+ decay of 38mK. The value of a is a sensitive probe of possible scalar
contributions to the weak interaction; a measurement to 0.1% precision would com-
plement the high-energy searches at accelerators. Although continued data collection
is planned to increase statistics with an improved geometry, TRINAT currently has
enough for a 0.3% measurement.
This thesis, after an overview of the theory and the correlation experiment, will
describe the design and characterization of the 0-telescope used by TRINAT to observe
the momentum of the emitted positron.
Chapter 2 is an introduction to the theory of f l decay. The tint section provides
a detailed description of Fermi's mode1 so that the readet can have an intuitive un-
derstanding of the decay. The more modern and general view of 0 decay is outlined
in the second section, and shows how the correlation parameter affects the decay. A number of books 11, 2, 31 were useù as general reierence guides For this section.
Chapter 3 is rneant to provide details specific to the correlation experiment. Details
of the method TRINAT is using to measure a are given in the first section. Following
this is an overview of magneto-optic traps and how TRINAT uses them in the exper-
iment. The last section gives a brief description of the nuclear detection system as
a whole, which consists of the 0-telescope contained in this thesis as well as a recoil
detector.
Chapter 4 describes in detail the P-telescope. The first section explains how the
the telescope as a whoIe was designeci and with what considerations. Section 2 is a
detaiied description of the double-sided silicon-strip detector, its characteristics and
the results obtained from the 38mK data. Following this is a similar section, this
time dedicated to the plastic scintillator. Section 4 presents on-line results of the
@-telexope as a whole and the final section gives a preliminary analysis of the ,&Ar coincidence spectra, although no attempt is made to calcdate a as it is outside the
CHAPTER 1 INTRODUCTION 3
scope of this thesis.
Chapter 5 is simply a summary of important results and suggestions for future endeavors.
,6 Decay and Fundamental S ymmet ries
,û decay has proven to be an invaluable tool in our study of nuclear
and particle physics since the discovery of radioactive decay at the turn
of the century. Early experiments observed only two decay products: the
recoiling daughter atom and a charge-conserving ',û ray,' which was soon
realized to be either an electron (P') or a positron (O+). The simple
kinematics of such a two-body decay wvould require that the P energy
spectrum be a peak corresponding to the energy released in the transition,
as in a decay. The fact that the emitted P ray was observed to have
a continuous energy spectrum prompted Pauli in 1931 to propose that
fl decay is a 3-body process, the extra product being a light (or even
massless), neutral particle that interacts very weakly with matter and so
escapes detection. Only three years later, but twenty years before it was
ever proven to exist, Fermi called this elusive particle a neutrino when he
incorporated it in his theory of @ decay. Fermi's mode1 provides us with a simple, intuitive understanding of /3 decay that, even to this day, remains
essentially unchanged.
2.1 The Fermi Mode1
A nucleus that undergoes decay converts one of its neutrons into a proton, or vice
versa, so that its nuclear charge, 2, changes by A l but the total number of nucleons,
TABLE 2.1: Possible foms for an interaction consistent with Lorentz invariance.
Type Scdar
Pseudo-scdar
Vector
Axial vector
'rensor
A, remains constant. The basic processes underlying these decays a t the nucteon level
are:
Operator
1
"Is
TP Yp75
?,,ru - 'Yv-Yp
n - + p + e - + F e p-dway
p - n + ef + Y, O+ decay
p + e- --+ n + v, orbital electron capture (E)
i n e n Fermi first proposai his theory [41, little else was known except that the
force inducing the decay was weak compared to the strong force binding the protons
and neutrons together to form the nucleus. This a l l o r d Fermi to use first order
perturbation theory to derive his Golden Rule, which gives the transition rate for any
suitably weak potential v,, (using natural units so that A = c = 1);
where M A = I,+j kt II, d3x is the matrix element of the interaction and p(Ef) is the density of states avaiiable to a final state of energy EI. The mathematical form
of FnL for weak interactions is not predicted by Fermi's theory, but there are only five
combinations of the y-matrices that are Lorentz covariant; these interaction t-vpes are
given in Table 2.1. Inspired by electromagnetism, Fermi guessed a vector form for the
interaction whose strengtht is characterized by a coupling constant, GF. The Feynman
diagram of Figure 2.1 is Fermi's mode1 of the P+ decay of a nucleus, AX, using the
more modern concepts of hadron and lepton currents. The following discussion will be restricted to p+ decay, but calculations for p- decay are the same except for the
direction of the currents.
The contact nature of the mode1 implies the matrix element is simply a contraction
of the two currents:
where the lepton current is simply given by
The hadron current is siightly more complicated because it must transform one
of the protons within the nucleus into a neutron. This involves isospin, a symmetry
of the strong interaction, which views protons and neutrons as two 'states' of the
same particle, the nucleon. The isospin of a nucleon is T = $ and the projectionst
2'3 correspond to the two distinct states, the proton (T3 = -I) 2 and the neutron
(T1 = fi). The SU(2) structure of isospin is the same as angular momentum so
there are ladder operators, r*IT 'ri,) = JT(T + 1) - T3 (T3 f 1) IT T3 1), which
transform between these nucleon states. The hadron current for P+ decay is:
where Ive use the r+ operator since the isospin is raised in $+ decay.
ClVe begin calculation of the hadron current by separating the wavefunction of the
decaying nucleon from that of the remaining nucleus:
and
Here .SI, and $&, represent the isospin components of the initiai and final state (over-
dl) nuclear wavefunctions. Treating the nucleons as stmctureless, Dirac particles, we
use wavefunctions of the fornt
+In partide phpics, the convention is reverseci with the proton in the TS = f i subtevel. t~otlowing the convention of Halzen and Martin [II.
FIGURE 2.1: Fermi's contact, vector-interaction model for the Bf decay f ; ~ - z- 1 AY + ef + ue.
where the x spinors carry the spin of the nucleons. The energy released in P decay is
much smaller than the nucleon masses; thus we can take the non-relativistic limit of
Equations (2.7) and (2.8). The overall nuclear wavefunctions are then:
,@,y - J?E; (O) e - i E p t ,+nucl @, E P - ~
Substituting these expressions in Equation (2.4)' the hadron current in the non-
relativistic limit becomes:
Note that in this limit, only the p = O (time) cornponeut of 7' contributes to the
matrk element. The spatial components, responsible for leaving the neutron in a spin
state difIerent from that of the proton, do not contribute in Fermi's vector model of
,B decay.
The leptons are naturaliy assigned Dirac wavefunctions, but h this case the non-
relativistic reduction is not valid. We only need the time component of the current
as for the hadrons, so the lepton current is
where for the sake of generality Ive do not assume a massless neutrino. fl decay
energies are on the order of a few MeV and typical nuclear sizes are a few femtometers;
therefore, over the nuclear volume, (p , + p,) . x (< 1. This ailows us to expand
the exponential in Equation (2.12) and to keep only the tirst term (the "allowed
approximation1'):
Physically, the allowed term corresponds to the leptons being created at x = O with
no net orbital angular momentum carried away; any change in nuclear spin must be
reflected in the aiignment of the spins of the leptons which can be either parallel
(S = 1) or anti-paratlel (S = O). For the Fermi decays we are considering, the lepton
spins must be anti-parallel since according to Equation (2.11), the hadron current
vanishes if X, $ X, so there can be no net change in the nuclear spin, 1. The lepton
spins can be parallel in B decay, but in this case the operator is axial-vector, which
Fermi did not consider; these are referred to as Gamow-Teller decays which have the
selection d e A i = O, f 1. If I" = Of for both the initial and final nuclei (where
r refers to their parity), A l = O but since no orbital angular momentum can be
transferred, it must be a 'pure Fermi' decay (no Gamow-Teller components).
The matriv element, Equation (2.2), of Fermi's fl+ decay c m now be evpressed
as:
where k/[F is the (Fermi) nvclear matrix element of this decay which accounts for the
effects of nuclear structure. Our choice of wavefunction normalization corresponds
to having 2E particies per unit volume, that is uh = vtv = 2E, which we use after
quaring the matrix element Equation (2.14):
This choice of normalization also makes calculating the density of states rather
simple: with 2E particles, the individual phase space available to a particle in the
momentum element p+d3p is simply (d3p/(2?r)3)/2E. On the basis of Equation (2.1)'
the differential decay rate of ZX 3 '-:Y + e+ + u. from rest is
with energy-rnomentum conservation ensured through the 64 function. The matrix
element contains the 'physics' of the interaction, but the observed spectrum shape
will be dominated by the final state phase space. Letting Eo represent the energy
released in the decay, substituting Equation (2.15) into Equation (2.16), and using
the relation p2dp = p 2 f d ~ for both the electron and the neutrino gives
Finally, we use the 6 function to integrate over dE, so that ive end up with the
decay rate:
The additional 'Fermi factor', F(Ee,Zt,R), accounts for Coulomb interactions be- tween the emitted positron and daughter nucleus. Fermi derived an analytic expres-
sion For this function when he b t proposed his theory 161:
where s = d m l 11 = *Ee/pe and Ris the nuclear radius taken to be 1.2 A ' / ~ fm. cr = e2/47r is the fine structure constant and 2' refers to the daughter nucleus. Due
to the difficulty of evaluating the complex I' function, Fermi used a non-relativistic
approximation but since we now have cornputers to aid us, Equation 2.19 is used in
calculations throughout t his t hesis.
Althougb Fermi's theory worked very well, there were still problems (specifically
with the decay of the kaon system) which prompted Lee and Yang [71 to question
whether parity was strictly conserved as required by a purely vector coupling, There
were, after ail, four other types of interactions (Table 2.1) which Fermi could choose
from, and some do not conserve parity. Madame Wu's experiments with polarized
60Co, as well as more accurate experiments that followed, indicate that parity is
mmirnally violated in weak interactions. The form of the weak operator as presently understood is V - A; it is the (equally large) axial-vector component which aIlows
Gamow-Teller decays.
2.2 A Generalized Interaction
Modern particle physics is based on quantum field theory in which the forces are
mediated by the exchange of particles, the quanta of the fielcl. The electromagnetic
force, for example, is mediated by the emission and absorption of its quantiaed fieid, the photon. For the weak force, there are three massive bosons: the W* which carry
one unit of electric charge and the Z0 which mediates weak neutral currents. The
weak coupling constant, g,, is related to Fermi's constant by
and gives the relative strength of the mak force. The mass of the W* has been mea-
sured (5) to be 80.41 f 0.10 GeV so that g, x 0.653 is greater than the corresponding
constant for the electromagnetic force, g, = 6 x 0.303. Indeed, the only reason
this force is 'weak' compared to that of electromagnetism is because of the large mass
of the propagator-
If we limit ourselves to pure Fermi decays (ones where there are no Gamow-Teller
components in the matrix element), then the axial vector component of the weak
interaction vanishes as mentioned earlier. SMilariy? the pseudo-scalar and tensor
components do not contribute. Consequently, we oniy have to consider vector and
scalar interactions; with only Lorentz invariance required, Figure 2.2 shows their
Feynman diagrams. The strength of each interaction type is reflected in the lepton
vertex parameters, CSpv and C&,V, which must be determined experimentally. The S
stands for a scalar current and the V for vector; the primed parameters allow for parity non-conservation in the lepton current, while tirne-reversal invariance is assured oniy
if al1 the parameters are real. We furthermore assume for simplicity that the scalar coupling constant is the sarne as the vector, and that the scalar boson's m a s is the
same as the charged W . In the limit that the momentum transfer, k, is small compared to the LV mass
(applicable to /3 decay), the propagaton reduce to constants: (vector) and 4 (scalar). It is for this reason that Fermi's contact approximation worked so well:
the mass of the W makes the interaction extremely short-ranged (on the order of
l/MI.v x 0.003 fm). The interactions depicted in the diagrams of Figure 2.2 are (with the same limits/approximations as in 52.1):
-i(gpu-k&/M$) (a) vector propagator: kl-hffv (b) scalar propagator:
FIGURE 2.2: General Feynman diagrams for the OC+ Of P+ decay of 5.X - ' - ;~+e++v. as mediated by a massive boson. The expressions for the propagators are given below for bath vector and scalar interactions.
Note that in the non-relativistic b i t , the hadron current is the same for vector
interactions as it is for scalar. The final matrix element is Mli = Mv +Ms which, after squaring and evaluating
the traces, yields three terms:
Ee EvC P e Pv.
IMrI2= 1 6 ~ p M W v ~ G [ ( I c s ~ + I c ~ I ~ ) (1 - =) -
M ~ M ~ + M ~ M , = ~GE,,E,,E.E,G~
-el (2.2,) 2%e(CvCS + C' Ch -
Insertion of these equations into Equation (2.16) and continuing the same calculation
yields the generalization of (2.18), the decay rate for vector and/or scalar interactions
where massive bosons propagate the force:
The decay rate of Fermi's mode1 is effectively renormalized by
and the other parameters are defined as:
in the Standard Model, Cs = CL = O and Cv = CL so that for O++ O+ decays,
it predicts that the p- v correlation parameter a = 1. If instead the weak interaction
is mediated by a scalar boson (Le. Cs = CS and Cv = C(I = O), then the correlation
parameter is equal to -1; in this manner, a precise measurement of a provides a very
sensitive test of possible scalar currents in weak interactions. A determination of a to
nrithin f0.5% would be complementary to more direct high-energy searches [8, 91. The
best meanirement to date is of 6 = i+4(:e,g) = 0.9989 f 0.0052 sys f 0.0039 sys [101 from a detailed measurement of the energy spectrum of P delayed protons from the
pure Fermi decay of 32Ar(Of, T = 2) to the lowest O f , T = 2 excited state of 32Cl.
The b, and 6, parameters are the amplitudes of the Fierz interference terms which
uise from the M ~ M S + MLMV cross terms in the matrù element; beU represents
an interference between the electron and neutrino wavefunctions within each of the
vector and scalar interactions. The bue and b,, terms are negligibly small (indeed,
possibly zero) since m, « Eue, but are included for completeness and to expose the
symmetry. A limit of IbelS 0.007 has been independently placed using measurements
of the Ft values of Of -4 Of decays [Ill. Jackson, Trieman and Wyld [121 were the first to calculate the general decay rate
Equation (2.24) for allowved ,8-decays (including mixed Fermi/Gamow-Teller transi-
tions). Their calcuIations assume a massless neutrino and so do not include the 6, and bew terms. They do, however, include Coulomb corrections [131 which adds a
LZm%%(CsC; PC +CkCg)/ [ICs12+(C~12+l~v12+IC~12] term to a and multiplies b. by
the relativistic factor d m .
The ,tl - v Correlation Experiment
The of decay of the isomeric state of potassium-38 is one of the a g
proximately twenty known cases of strong O++ O+ decays that Nature
has ofTered us and, as an alkali atom, is perfectly suited for neutral atom
traps. The decay 38mK -+" B + ee+ + ut from a trap provides us with
an excellent laboratory with which to search for scalar currents.
This chapter is rneaat to provide some details specific to TRINAT'S 0-v correlation experiment . The first section explains how the observables of
the decay in the back-t+back geometry can be used to determine a. Fol- loiving this is a brief introduction to neutral atom traps, TRIN.W'S double-
trap system and how ive get our radioactive potassium. An overview of
our detection system as a whole is given in the Cinal section which pri-
marily deds with the recoil detector since the P-telescope is the focus of
Chapter 4.
The decay schemes for both the isomer and ground state ildl are given in Figure 3.1.
The 38mK decay is a pure Fermi transition with a branching ratio measured 1151 to
be > 99.998% and a Q-valuet of 5.02234(12) MeV [16l. The 3BK decay is also given
because I s~c ' s mass analyzer (see 53.2.2) canot isomerically differentiate between
the ground state and 38mK, making us susceptible to the ground state7s .y background.
In what follows, let us consider observing the decay fiom a point-like source in
tFor /3+ decay, Q = E, - me.
3.1 USING ""K TO MEASURE a 15
3 8 m ~ ) = 5.02234(12) MeV QP+ (19 E, = 130.4 keV
F~CURE 3.1: Nuciear (log f t = 3.5) is by far pf decay feeds the Zf
energy ievels for 38mK. The superallowed Pf decay the most favoured branch for the isorner. The ground state's state resdting in a 2.17 MeV y.
the back-to-back geometry where the positron is emitted opposite the recoil. The
neutrino may be emitted either (a) pardel to the P or (b) parallel to the recoil. The
kinematics of the decay will be different for the two cases as shown in Figure 3.2; this
is a plot of the recoii's tirneof-flight (T'OF) against the positron's (kinetic) energy.
The recoil TOF is the observable used in the correlation experiment (along with
the position of the event) to determine the recoil momentum, so we use the TOF here for consistency. The TOF is uearly independent of P energy in the fast branch
(case (a)) because the recoil velocity is not seriously affected by how the leptons
share the rest of the available energy: the recoil velocity is [E, - Ee(l - ve)] /MAr,
which is a constant to order (1 - -u,). In the slow branch (case (b)), it is the recoil
which is sharing momentum with the neutrino, and in this case the kinematics yield
v~~ = [Ee(l + ve) - Eo] /hlh. The 0's energy in the asymptotic limit of the slow
branch is defined by the relation: Te + pe = Q. The generalized decay rate, Equation (2.24), shows how the population of the trvo
branches depends on the form of the weak propagator: if purely vector (as in the
FIGURE 3.2: Recoil timeof-ûight vs. B energy for Aro in the back-to-back geometry. For this discussion, we take the neutrino to always be left-handed with helicity = -1 as shown and, depending on the interaction type, the positron is preferentiaiiy emitted with either equal (scdar) or opposite (vector) helicity. The slow and fast branches are either enhanceci or suppressed since the total lepton spin must add up to zero.
6
T 5 - a u
2 4 - M .d
3 - I
'H O l 2 - a, E G 1 -
1 l I l
Vector - (nrpprmstad by ( 1 -
-
-
- "Fast" branch
C
=Ar -. *e+ -. - ."*
*e+ -4
Vector Scalar (mhanced by ( 1 - v&J) ( ~ p p t e s s a d by (1 - Y/=))
3.1 USING 38MK TO MEASURE a 17
Standard Model), a = +l so that the fast branch (je IJ, = cos 88, = 1) is enhancedt
by 1 +v, 2, whereas scalar interactions (a = -1) are suppressedt by 1 - ve 2 0. In the slow branch, cos Op, = -1 so the situation is reversed, and this time it is the
scalar currents which are enhanced (with the vector suppressed). In this way, the
back-to-back geometry is the most sensitive geometry with which to search for weak
scalar bosons.
The enhancement and suppression of the two branches c m be understood physi-
cally in terms of the outgoing lepton helicities. Helicity is defined as the projection of
a particle's spin, E, dong its direction of motion. The projection operators, i(1 j$),
can be used to pick out a particle's helicity, i.e. for the (presently assumed massless)
neutrino, we can project out its left- and right-handed components:
These chiral states are helicity eigenstates with eigenvalues +1 (right) or -1 (left).
For particles with mass, however, this is only true in the relativistic limit since 1171,
where $+ represents a particle (u) and $L an anti-particle (v). Note that for rnmless
particles only, helicity and handedness can be used interchangeably; a left-handed
neutrino will always have negative helicity. A left-handed massive particle on the
other hand predominantly carries negative helicity, but a component that goes like
(1 - E) in the relativistic limit has the 'wrong' (positive) helicity.
To see how this affects the 3smK decay, Iet us for the sake of clarity take the
neutrino to be only left-handed so that the scalar current of Figure 2.2 (on page 11)
then becomes 3v ," (cS + CSy5)ve. The 7= of t he projection operator anti-cornmutes
with the 7' in the adjoint, so zL = ~ i ( l +y) and the current can be seen to only
+Neglecting the negiigible (but conceivably possible) contribution Eiom the b., term.
3.1 USING ""K TO MEASURE U 18
couple left-handed neutrinos to left-handed positrons:
Similarly, right-handed neutrinos only couple to right-handed positrons so that for
scalar interactions, the leptons are emitted with the same handedness. In the limit
of zero mas, their belicities are equal; the positron may have the wrong helicity, but
this is suppressed because it is relativistic and so the approximation in Equation (3.2)
is very good.
The vector current of Figure 2.2 contains a yfi at the vertex, which anti-commutes
with the y5 and so changes the positron projection. A left-handed neutrino in this
case couples only to a right-handed positron (and a right-handed v to a left-handed
ef). Vector interactions, therefore, differ from scalar interactions because the leptons
are emitted with opposite helicities.
As mentioned earlier, the spins of the leptons in a 0'- O' decay must add up to
zero. Retuming to Figure 3.2 (which depicts the lepton spins as fat arrows for each of
the four decay possibilities), it is easy to see how the fast and slow branches are affected
by the type of interaction. In the fast branch, the right-handed positron (resulting
tiom a Standard Mode1 vector boson propagator) naturally has a spin opposite to that
of the neutrino's, and so is allowed. In the scalar case, the dominant spin of the left-
handed positron is aligned with the neutrino's which is forbidden; the positron would
have to have the wrong helicity to be in the fast branch (depicted as the shorter of
the two electron spin arrows). Similarly, the slow branch favours the leptons canying
equal helicity since in this case the neutrino's momentum is opposite the positron's.
TRINAT wiil determine a by simultaneously fitting the recoil TOF spectra for a
number of p energy bins to detailed Monte Carlo simulations. The shape of these
spectra will depend on the ratio of slow to fast branch events and hence a. The B - v
correlation experiment and the anaiysis of its results is largely a part of A. Gorelov's
Ph-D. and so the interested reader is referred to his thesis [la].
3.2 Trapping Techniques
Ever since the first successful experiment where neutral atoms were trapped using
lasers and magnetic fields [191, the use of magneto-optical traps (MOTS) has become
widespread in many fields of physics. The ability to observe radioactive decays from
such traps opens up new possibilities in precision B decay experiments. The MOT
( i ) confines the decays to occur within a compact (= 1 mm3) trap volume, ( i i ) cools
the atoms to temperatures at or below the mK level, and (iii) allows distortion-
free detection of the daughter particles' momenta because of the negligible source
thickness. For this section alone, we use SI units instead of natural units.
3.2.1 The Neutra1 Atom Trap
A unique feature of magneto-optic traps is that their force is dependent upon the
atom's position as well its velocity; much like a spring, the positional force retums
atoms back to a common centre while the latter adds a damping effect, as if the spring
was submerged in a viscous liquid. A laser field generates the velocity-dependent
force which cools the atoms in an optical molasses, while a polarized laser field and
an applied magnetic field adds the positional dependence needed to actually confine
them.
InitiaMy, if one considers how small the incident photon's rnomenturn, hk, is com-
pared to that of thermal atoms, a trap based on the light forces From a laser beam
seems futile. For example, the SII2 4 P3/? (4) transition in 38mK bas a wavelength
of X = 766.5 nm so that Pphor 1.5 eV/c while the atom's thermal momentum is
typically 45 keV/c. In addition, when the atom decays from the excited state via
stimulated emission, this small momentum kick is nullified since the atom will recoil
by -hk if the photon is emitted coherently. Spontaneous emission, on the other hand,
emits photons homogeneously into 47r so that, at least on average, there will be a net
momentum transfer in the direction of the incident photon. Clearly, the atoms must
absorb and spontaneously emit many photons if one hopes to optically trap them. atoms are well-suited for neutral atom traps because they have a simple
electronic configuration; the nsL valence electron alone determines the âne and hyper-
fine structures of the atom since al1 other electrons are in closed (noble gas) shells.
The atomic angular momentum, J = S + L, and nuclear spin, 1, of an atom are a
coupled system, so the good quantum states are IF, mF) where F = J + I. The tran-
sition h m the F = 1 + 4 ground (Si/2) state to the FI = I + excited (P3p) state
cannot decay back into the other F = I - $ ground state since the photon carcies one
unit of angular momentum; the atom must retum to the same F = I + !j ground state
where it started. This 'cycling transition' allows an atom to absorb many photons
from the same laser beam so that the light forces can build up and significantly affect
the atom's motion. The cycling is not perfect due to off-resonant transitions and finite
linewidths, and so atoms rnay be optically pumped into the F = I - state where
they will no longer be in resonance with the laser light. Atorns in this Ldark' state are
not trapped, so to transfer tliem back into the cycling transition, a 'repurnping' laser
is tuned to the F = I - 4 -, F = I + & transition.
It is worth noting at this time that for 38mK, I = O which simplifies the atomic
energy levels; there is no hyperfine structure in an atom with no nuclear spin. In this
case, there is no need for the repumping laser.
Doppler Cooling
Consider the effect on an alkali atom that is placed within a laser field generated
by two identical, counter-propagating laser bearns detuned A MHz below the atornic
resonance, WA. Thermal atoms will have a Mawellian velocity distribution in al1
directions and, due to the Doppler effect, will be affected by each beam differently.
An atom moving collinear with one of the laser beams d l see the light red-shifted
farther below resonance by a factor uatom/c while the counter-propagating beam will
be blue-shifted closer to resonance. The atom wiil preferentially absorb photons
from the laser beam that is against the atom's direction of motion and, with each
absorption, the atom's momentum is reduced. It is in this way that the atorns are
cooled. It can be shown [201 that the force in an opticai molasses is proportionai to the
velocity if ( i ) the Doppler shift isn't too large and (ii) the laser intensity, IL, isn't
strong enough to induce stimulated emission. Letting w t = U A - A represent the
frequency of the laser light, the Doppler force is:
where I' = 1/r is the transition linewidth and I, = " is the saturation intensity.
The transition linewidth ends up being this force's limiting factor; in a simple
mode1 [201, the coldest temperature attainable is TD = hï/2ks where kg is Boltz-
mann's constant. In the case of 38mK for which T = 26 as, this 'Doppler limit' is
150 pK.
The MagnetwOptic Trap
An atom with magnetic moment p subject to a magnetic field, B, will remove the
r n ~ degeneracy and Zeeman split the 1 F, .mF) energy levels according to
where g~ is the (atomic) Landé g-factor and p~ = eh/2m, is the Bohr magneton. If we
apply a linear magnetic field, Bz = Boz, to atoms in an optical molasses (Figure 3.3
depicts this for the 1-D case), the Zeeman effect will provide us with a position-
dependent force: the energy level shifts (hence the transition frequency and scattering
rate) will be proportional to the distance from z = O. An atom at z > O d l have
its 1; 1) - 1 9) transition shifted closer to the laser frequency. The two counter- propagating beams are distinguished by their polarizations; the AmF = -1 transition
can only be driven by the a- beam, so the atom feels a net force acting towards the
left. Similarly, atoms at L < O will preferentially absorb the a+ light through the
I$ +)- 1; 1) transition and be pushed back towards centre. If the laser field does
not have perfect polarization, the atom can absorb photons from the wrong beam and
will be heated rather than cooled. Thus, the quality of the circular polarization of
the laser light is very important.
The magneto-optic force is a combination of the Doppler force with effects induced
by the Zeeman shift. The trap centre is defineci at t = O where the magnetic field
changes sign. If we let = h + ,8t where Pr represents the Zeeman effect on the
(a) z, B c O (b) Z = B = O (c) z. B > O
h h 4 4 \ $ 1 1 \ \ l 1 \ 1 I l II 1 \
1 \ A l \ l \ 1 \ l \ l \ I $ 1
\ 1 \ 1 \ 1 \ 1 A& \ l A m~ = -L 1
2 2 Force no net Force - force f-
FIGURE 3.3: Atomic energy levels and Zeeman shifts of 38mK in a one dimensional magneteoptic trap. The shift of the energy levels depends linearly on the magnetic field (for weak fields) giving a positional dependent force. The trapping arises because of preferential scattering off the laser that is shifted closer to resonance, depicted by the solid Iines arising from a+ light in (a) and a- in (c); the force in either case pushes atoms towards 2 =O. Both laser beams are equdy off resonant at the trap centre (b) so neither is preferentiaily absorbed and the atom feels no net force.
transition frequency, then the overall magneto-optic force is [211
It can be shown that (221, for small detunings, this force is proportional to so that
the trapped atorns behave like a simple, damped harmonic oscillator.
The generalization to three dimensions requires a quadrupole magnetic field that
is zero at the centre and from there, increases linearly with distance; this is generally
accomplished using two coils in the anti-Helmholtz configuration. With six laser
beams oriented along the axes as depicted in Figure 3.4, the forces in the ID case are
present no matter which direction an atom in the trapping volume is travelling. Typically, MOTS are not very deep and so considerable effort is needed to effi-
ciently load them. The rnost popular rnethod is known as the vapour-ce11 MOT [191
and it uses the fact that a MOT can capture the low-energy tail of the Maxwell-
Boltzmann distribution of velocities, The ce11 which defines the trapping volume can
be specially coated (231 so that atoms tend to bounce off the wall rather than (perma-
nently) chemisorbing to it. When they bounce (physisorb), the atoms ce-thermalize
(repopulating the whole ;Llauwell-Boltzmann distribution) and are able to be trapped
by the MOT again. The many repeated opportunities for capture allow this method
of loading the MOT to have efficiencies on the order of several percent.
3.2.2 T R ~ F ' S Radioactive Potassium Source
TRIUMF bas long been interested in developing radioactive ion beams (RIBs) because they offer the ability to study nuclei atvay from the valley of stability. The copious
production of exotic nuclei delivered with low kinetic energy has already been used
by TRINAT as well as an experiment measuring the transition probabilities of su-
perallowed decays which will test the conserved vector current hypothesis [241; future
planned experiments at T R I ~ F ' S RIB Rclude magnetic moment measurements using
the Low Temperature Nuclear Orientation facility [251 as weU as studies using a po-
larized 'Li beam [261. b1edium-energy (0.15-1.5 MeV/u; A<30) EUBs are important
for nuclear astrophysics because theoretical calculations of stellar processes require
measurements of certain nuclear crosssections that affect key steiiar cycles [27, 28).
FIGURE 3.4: Schematic diagram of a vapour-ce11 MOT with an ion beam catcher and neutraiizer. SAC provides us with a potassium ion beam which is neutraiized in the hot Zr cone. The atoms then diffuse out into the vapour-celi MOT and the ones in the low-energy t d of the velocity distribution are able to be trapped in the MOT. The quartz cube defines the trapping volume and is coated with SC-77 Dryfilm [231 so as to maximize the number of bounces off the w d s and increase the trapping efficiency.
FIGURE 3.5: The ISAC radioactive beam facility at TRIULIF. The 500 MeV positrons bombard a Ca0 target, producing a variety of isotopes; a potassium beam is extracted and filtered through a mass spectrometer before being transported to the TRINAT laboratory.
A test facility, TISOL 129, 301, based on the ISOL facility at CERN-ISOLDE (311
has been operational since 1987 to provide low energy RIBs as tell as to develop
a robust target-ion source. The 200 - 500 MeV protons from TRIULIF'S high inten-
sity (150 PA) cyclotron provide an excellent production beam with which to bom-
bard thick targets, and TISOL has proven that a wide variety of radioactive nuclei
can be efficiently produced (500 MeV protons; 1 pA current). Ion beams were
extracted from the target at 6 - 12 keV, and transported to a magnetic andyzer
(mas resolution M/AM = 1600) that had a bend angle of 90" to provide a clean,
tunable beam of radioactive isotopes to the experimental area. TRINAT was able to
successfuily couple TISOL'S RIB to a MOT, continuously trapping 2 000 atoms of 3TK (tLI2 = 1.23 s) and 5000 of 38mK. From this, the viability of the correlation experi-
ment was proven and measurements were made of the isotope shifts [321 in potassium
as well as charge-state distributions of daughter atoms in ,û decay [33, 341. The success of TISOL and the experience gained in developing it has enabled
Isotope Intensity
38mK 2 x 107 ions/sec
38K (bkgd) 5 x 10' ions/sec
TABLE 3.1: [SAC radioactive beam intensities.
TRIUMF to successfully build a large scale radioactive beam facility: ISAC (the Isotope
Separator and ACcelerator). A schematic diagram of the new beam facility is given in
Figure 3.5. The general idea is the same as for TISOL, but the new surface ion source
can handle more intense production beams (1 pA in June, 1999; currently 10 pA and
up to 100 pA in the future) and the ions cm be extracted between 12-60 keV. ISAC'S mass analyzer consists of a low-resolution pre-separator followed by the former Chalk
River mass separator 1351 which has a mass resolution of 1 5000. The isotopically
selected RIB is then transported to the various experimental areas; in the case of TRINAT, the beam is deposited into a conical zirconium foi1 as indicated in Figure 3.4.
ISAC'S first radioactive beams, 37*38mK, were produced on November 30, 1998. By
TRINAT'S June 1999 run, ISAC had already demonstrated the ability to produce these
beams with intensities comparable to those attained at TISOL (see Table 3.1). The vapour-ce11 MOT of Figure 3.4 aiiows an efficient loading of atoms into the trap, but
is completely incompatible with obseMng the recoil of the decays. Efficient loading
of an open MOT can be obtained by carefully transporting atoms already trapped in
the vapour-ce11 MOT; this system is outhed in the next section.
3.2.3 TRINAT'S Double MOT System
in order to reduce backgrounds as well as to isomerically differentiate between 38K and 38mK, a double-MOT system is utilized (see Figure 3.6). The first 'collection
trap' is a vapour-cell MOT that traps the potassium atoms once they diffuse out of
the hot Zr neutraiizer (which is based on the Los Alamos scheme [361). The laser
linewidth is small enough to resolve the difiecent fine structure of the isomer and
ground state, so only the isomer is trapped. The cold, trapped 38mK atoms at the
centre of the collection trap are then pushed with a smali (w 1 d) pulsed Iaser
FIGURE 3.6: Schematic diagram of TRINAT'S B - u correlation experiment. The ion beam from ISAC is deposited and neutralized in a Zr foi1 and then trapped in the first 'collection' trap. The cold, trapped atoms are then transferred to the 2" trap where the decay is observed using a micro-channel plate (recoils) and the Ptelescope (positrons).
beam to a second 'detection trap.' The push beam, airned a few millimeters above
the 2" MOT so as not to interfere with any atorns already trapped there, generates
a low-energy (vat,, 40 m/s) atomic beam of 38mK that can be directly captured in
the detection trap. The large separation of the two traps reduces the probability of
a thermal 38K randody entering the detection charnber, and so provides a cleaner
environment from which to observe the decay.
As the atoms are being transferred they expand radially, so the efficiency of the
transfer wiil rapidly decrease as the inter-trap distance is increased. In order to be
able to separate the detection trap 75 cm (enough to add approximately 15 cm of
lead shielding) without a great loss of atoms, two 2D MOTS ('atomic Funnels') have
been employed t o compress them back dong the push beam axis. The detection trap
typically catches 75% of the atoms pushed from the lSt trap. The efficiency of the
system as a whole, including capture and transfer, is 5 x 104 atoms trapped per ion
incident h t o the neutrdier.
For a more thorough review of TRINAT'S double-MOT system and the details
regarding the transfer, the interested reader is referred to [371.
3.3 Nuclear Detection System
The P - v correlation experiment requires the detection of the recoiling Ar atom and
the emitted P+ and, as rnentioned earlier, is most sensitive to scalar contributions in
the back-to-back geometry. The recoil detector has a nominal active area of 2.6 cm in
diameter and is placed -6.28 cm along the 2 direction from the center of the chamber
(which roughly corresponds to the detection trap center). The ,L-telescope is on the
opposite side and consists of a 1.2 x 1.2 cm2 double-sided silicon strip (AE) detector
at z = +6.76 cm backed by a large plastic scintillator (E) detector. The solid angle
subtended by the recoil and ,L detectors are 0.14 and 0.13 sr respectively.
The MCP purchased from Galileo [381 and depicted in Figure 3.7(a) is a thin array of tens of thousands of tiny cylindrical lead glass channels (10 - 15 Fm in diameter).
Each of the specially formulated micro-channels acts as a miniature electron muiti-
plier tube. An ion incident on one of the channels will generate secondary electrons
which are accelerated down the channel wall and generate further secondary electrons, resulting in a cascade which yields amplifications up to a few times IO4. The applied
high voltage bias provides the electric field along the length of the channel, and s u p
plies the electrons needed for the avalanche. The channels are oriented N 11" with
respect to normal of the MCP surface in arder to minimize variations in the detector's
efficiency as a function of the incident particle's direction which is related to Op-,.
TRINAT'S recoil detector is a combination of three 600 pm thick MCPs in the 2-stack
configuration (see Figure 3.7(b)) with an inter-plate separation of 150 Pm. The signal
from the 2-stack, which has an amplification up to IO", is registered by a resistive
anode which has four separate readouts; the relative distribution of charge between
the readouts allows the position of the recoil to be determined to f0.25 mm.
As T R ~ A T is determinhg the recoil momentum from the position of the hit in the
MCP and its the-of-flight relative to the plastic scintillator, the MCP is operated at
saturation (w 1 keV bias/plate) to make its efficiency as insensitive to the incident
ion's energy as possible. The timing resolution of the MCP is excellent (characteris-
tically hundreds of picoseconds), and so it is not surprishg that the relative timing
between the recoil and ,L detectors is limited by the scintillator's timing (see 8i1.3.2).
The MCP efficiency for Ar recoils can have a strong dependence on the atom's
(a) A micro-channel plate (b) Operation in the 2-stack configuration
FIGURE 3.7: Schematic diagram of an MCP detector and the arrangement used by TRINAT. The distribution of charge of the four resistive anode readouts is used to determine where the event occured.
energy; over energies of O -450 eV, the MCP efficiency is not known. This fact makes
using the neutral Ar recoils for the correlation experiment very difficult because the
slow branch covers a large recoil energy range. For decays where the daughter Ar
is a positive ion, however, this source of systematic error can be greatly reduced;
TRINAT uses a uniform electric field to accelerate the charged Ar recoils up to energies
where the I\1CPts efficiency is known [391 to be relatively Bat SE^[^^ < 2% between
EAr = 5.3 - 5.6 keV).
Additionally, this field separates the different charge states in TOF and increases
the number of detected recoils by focusing Ar ions that wouid have othemise missed
the MCP. Collection of the complete angular distribution is obtained for charge states
greater than +3 with our present field of -829 V/cm.
The double-sided silicon-strip detector (DSSSD) provides both position informa-
tion of the ,8 as well as a coincidence condition to be used with the scintillator to
d u c e background (non+) events. The energy deposited in the scintillator (and, to a
lesser extend, in the DSSSD) together with the position of the hit in the DSSSD pr+
vides us with a measurement of the p's momentum. Therefore, the energy resolution
of the scintillator needs to be as good as possible for the ,8+ momentum reading. With
the recoii and positron's momenta both measured event-by-event, the momentum of
the neutrino can be deduced; in fact the kinematics of the decay are overdetermined, and so we are able to check systematic errors in our experiment.
The Positron Detector
This chapter will first go through the design of the scintillator and will
then go on to the characterization of the A E and the E detectors. The &telescope as a whole is discussed in the h a 1 section where results from
the April/May 1999 run pertinent to the - u correlation experiment will
be presented.
4.1 Design and Construction
For the 0-v correlation experiment to succeed, the emitted positron must be detected with good energy resolution and have good timing relative to the recoil MCP detector.
In general, the energy of P-particles is difficult to measure accurately because they
are relativistic over the 1 - 5 MeV region of interest; because of their small mass,
they scatter easily, into large angles and emit bremsstrahlung radiation which may
escape detection. The situation is worse for Of's since one must also contend with
the possible detection of the annihilation radiation. An added complication is the
possibility of annihilation-in-flight, in which case E, > me and it rnay add to the
low-energy t ail.
4.1.1 Design Considerations
Early T ~ A T expeciments utilized a double-Si(Li) detector in conjunction with a
double-sided Si-strip detector (DSSSD) to fom the E and AE detectors of a 0- telescope, which has been discussed in detail elsewhere [401. The Si(Li) semiconductor
devices naturally have evcellent energy resolution (FWHh,I = 60 - 120 keV over the
4.1 DESIGN AND CONSTRUCTION 32
region of interest) as well as iinear energy responses, but suffer from a number of
serious drawbacks: they have a large low-energy tail from bremsstrahlung and the
large probability of backscattering off the high-2 material; a systematic error for p's
that penetrate the first Si(Li) due to the necessary dead-layer between the detectors;
and its overall change in response (the peak of the response function at 2.5 MeV
is N 3x larger than at 5 MeV) 1401. Simulations of the correlation experiment 1411
compared B-telescopes where the E detector was (1) a double-Si(Li), (2) a 0 4 cm x
4 cm plastic scintillator, (3) a 07.5 cm x 5.1 cm plastic scintillator, and (4) a perfect
detector (6-function response). The simulations used measured response functions from [401 for (1) and (2); the pararnetrized responses of Clifford et. ai's telescope have
been published [42, 431 and were used for (3). The results indicated better sensitivity
to the correlation parameter if a plastic scintillator similar to the Clifford design ivas
used instead of the Si(Li).
4.1.2 GEANT Simulations
In order to optimize the plastic scintillator for TRINAT'S B - v experiment, Monte
Carlo simulations were performed for various geometries using the GEANT detector
description and simulation toot (441. The goal was to design a scintillator that is
insensitive to the entrance angle/position of the 0s upon the DSSSD.
The initiai geometry was extremely simple: a piece of silicon, representing the
DSSSD, was placed in front of a cylindrical scintillator, both 'magically' suspended
in space. The sizes of scintillator considered ranged fiom 4 - 6.5 cm in diameter
and lengths between 3.5 - 6 cm, while the strip detector was constrained to be
2.4 cm x 2.4 cm x 491 pm (the size of the existing detector). The simulations helped
to determine the best compromise between:
1. Too large a scintillator - Compton scattering of the annihilation radiation
adds a high-energy tail (the 'Compton toe') tbat approximately scales with the
volume of the detector.
2. Too small a scintiilator - the positrons may escape the detector before annihi-
lating thereby adding to the low-energy tail; also the likelihood of bremsstrah-
lung radiation escaping increases.
4.1 DESIGN AND CONSTRUCTION 33
O NO0 2000 3000 4000 5000 6000 Tri. (dashed) or Tc (solid) [keV]
FIGURE 4.1: GEANT simulations of response functions for a 06.5 cm x 5.5 cm plastic scintillator for 1,2,. . . ,5 hieV positrons. The solid line has a resolution Eunction folded in and the DSSSD's energy added back in, while the rlaslird Line represents a perfect energy reading which requires a DSSSD coincidence, but the energy is not added to the scintiiiator's-
Simulations of response functions for a plastic scintillator are depicted in Figure 4.1
for TB = 1,2,. . . , 5 MeV. The high-energy tail is due to Compton summing of the annihilation radiation and is most prominent for 1 MeV positrons (5A.3). The low-
energy tail contains events where the positrons escaped the detector before stopping, but also adding to this tail are cases where bremsstrahlung (radiative) energy lasses
escape the detector (sA.2) and/or the positron annihilates before depositing al1 of its
kinetic energy and the y quanta escape (5A.3).
The optimal geometry
Figure 4.2 shows the percentage of positrons that fired the AE detector but, due to
multiple scattering, missed the scintiliator and therefore represent lost events. The
plot on the leFt has the AE placed 1.0 cm in Front of the E detector while the plot
on the right is the case where it is only 0.3 cm away. The effects of backscattering
4.1 DESIGN AND CONSTRUCTION 34
+ backscatter off DSSSD
FIGURE 4.2: Monte Carlo design simulation of the fraction of positrons that mdti- ply scattered in the DSSSD and consequently àid not interact with the scintiiiator. The effect depends on the diameter of the scintiiiator if the DSSSD-scintillator spacing is 10 mm (left), but is neariy independent if this spacing is only 3 mm (right). The 7.6 cm x 5.1 cm design, based on Clifford's dimensions [42], was simulated for cornparison.
directly off of the DSSSD is, of course, independent of its position and so the same in both cases. The original DSSSD mount used with the Si(Li) was such that 2.5 cm
was the closest the AE could be to the Front face of the scintillator. The results
of Figure 4.2 indicate that almost al1 of the positrons are fully contained, virtually
regardless of the diameter of the scintillator, if the DSSSD is placed close enough to
the Çont face of the scintillator. This motivated the design and implementation of a
new mount (451 which enabled us to reduce the spacing to 0.292 cm (see also $4.1.3).
A major factor in determinhg the optimal diameter was the requirement that
al1 of the positrons corne to rest and annihilate within the plastic's volume. In the
experiment, the most energetic positrons will have the largest range and so simulations
of 5 MeV positrons were performed Nith a very large scintiiiator to see what the
required length and diameter must be to contain them dl. Figure 4.3(a) shows the
4.1 DESIGN AND CONSTRUCTION 35
O 1 i s i s f i l a i podUon 01 annihiiatlon [cm]
FIGURE 4.3: Monte Carlo design simulations of a pencil beam of 5 MeV positrons depicting the radius (a) and depth (b) of where in the scintillator the annihiLation occurred. The dashed line in (a) represents the radius of the scintiilator used throughout this thesis.
radius at which the positrons annihilated in the plastic. This aspect of the scintillator
design was based primarily upon this result, but other factors, such as the total
number of accepted events and the relative percentage in the (high- and low-energy)
tails, also supported a larger radius. The dashed line in the figure depicts the final
scintillator's size and one can see that virtually al1 of the positrons are cantained. We would have considered a slightly larger diameter, hoivever mechanicd constraints
due to the existing vacuum chambers limit us to the present design with a 6.5 cm
diameter.
The penetration of the 0s along the length of the scintillator (Figure 4.3(b)) is
weii contained below 2.75 cm. The original analysis scheme for the correlation ex-
perirnent, which involved excluding kinematically forbidden events, was thought to
be very insensitive to the ce-absotption of annihilation radiation, and so any extra
volume was not a major concern. The low-energy tail on the other hand, which was
the instigating factor for switching from the Si(Li) to a plastic, is a potentialiy large
source of systematic error, Therefore, we ordered and (at least for the present t h e )
have continued using a 06.5 cm x 5.5 cm piece of scintillator, saving the option of later cutting the length to = 2.75 cm.
The scintillator used throughout this thesis is 5.5 cm in length, and therefore the
Compton toe of the response function is larger than necessary; dthough we will see
4.1 DESIGN AND CONSTRUCTION 36
that the present scintillator suffices at this tirne, the existing scintillator should at
some point be compared to one whose depth is only 2.75 cm. If the light collection is
not seriously compromised or if the final analysis scheme does prove to be sensitive
to the re-absorption, the B-telescope should be upgraded with the srnaller piece of
plastic.
The P window
Some of the P-telescope's components (e.g. the scintillator's wrapping) are not com-
patible with the ultra-high vacuum needed by a magneto-optical trap and so it is
housed in a separate ('P-telescope vacuum') chamber as depicted in figure 3.6. The
front face of this chamber is 7 mm thick stainless steel with a 038 mm 'window' the
P particle can enter through to suppress the observation of activity from atoms that
do not decay from the trap. The positrons Coulomb scatter in the foil as they traverse this necessary dead
layer and, if the positron scatters into a large angle, the deduced positron momentum (calculated by the position of the subsequent hit in the DSSSD) would not be cor-
rect. Originally, the p window was a 0.025 mm thick stainless steel foil (2sbt a 26.5);
however, the characteristic scattering angle [461 that defines the cone the ps scatter
into scales with z~'~/$, so the steel is clearly not the optimal material. The scat-
tering angle should be cut in half by using a lower Z material like a commercially
available [47] beryllium foil (type IF1, 2 = 4-14).
In order to see if the position improvements warranted purchashg the relatively
expensive beryllium foil, GEANT simulations were performed to compare the beryllium
(0.127 mm thick) with the stainless steel. Low energy ps have the greater probability
of scattering into large angles, so a pencil beam of 1 MeV positrons was incident
normally on the centre of the P window in the simulations to see the position deviations
caused by the two types of foiis. The geometry in the MC is the same as for the actual
telescope used in the April/May 1999 run, which has the front of the DSSSD 2.9 mm
behind the 0 window. Figure 4.4 shows the radial position of where the positrons hit
the AE after traversing the two types of foiis. The distributions are approximately
Gaussian over the peak (fiom small angle scattering) but have more pronounced tails
arising fiom large angle (Rutherford) scattering. The widths of the distributions are
- Be - -%.Steel
-lo -5 O 5 DSSSD position [mm]
F~CURE 4.4: Monte Carlo design simulations of multiple scattering effects due to the p window. The soiid iine depicts a pencil beam of 1 MeV positrons entering the DSSSD after traversing a 0.127 mm beryllium foil while the tliistictl simulates 0.025 mm stainless steel. Both foils are placed 2.9 mm in front of the DSSSD.
0.46 and 0.97 mm for the beryllium and stainless steel respectively. For the stainless
steel foil, then, a p originally incident towvaràs a given DSSSD strip wvill only have lu
of this distribution firing that strip, whereas 2a will fire the correct one For beryllium.
This factor of two improvement is decidedly worth the effort, and so the ,û window
was changed to the 0.127 mm thick beryllium.
4.1.3 The Final Design of the Telescope
The Telescope Assembly
The realization of the plastic scintillator and of the P-telescope as a whole is depicted
in Figure 4.5. The scintillator, a 06.5 cm x 5.5 cm long BC408 plastic purchased
from Bicron [481, is optically coupled to a plexiglass light guide which in turn is
coupled to a Philips 4312/B 12-stage photomultiplier tube (PMT). Both the plastic
and the light pipe are mapped with a d a s e reflector, Teflon, to increase collection
of the scintillation light as discussed in 84.3.1. The scintillation light produced in
the BC408 (A,, = 425 nm) overlaps the maximum of the bialkali photocathode's
response (400 - 450 nm; 20% quantum efficiency). The Philips PMT was chosen
4.1 DESIGN AND CONSTRUCTION 38
Date: Mardi 30,1999
FIGURE 4.5: Schematic diagram of the P-telescope assembly (to scalej. The major components are described in the text.
because it was specified to have a good hybrid of energy and timing charactecistics. A transistorized voltage divider assembly (Philips VD123K) was used to minimize gain
fluctuations arising from variations of the dynode voltages.
The /3-telescope's chamber separates the trapping region's 2 x IO-^'' Torr ultra-
high vacuum from the telescope's poorer vacuum of typically 5 x 10-'' Torr. A conflat
fiange was modified by adding two vacuum pumping ports and four 16pin electronic
feedthroughs for the 48 DSSSD strips. The vacuum is maintained by an O-ring which
seals the light guide to the fiange using a specially designed clamp.
The PMT is surrounded by 0.2 mm thick p-metal which, in addition to protecting
the PMT from magnetic fields, also serves to make the system light-tight. A blue
LEDt Mth a nominal wavelength of 450 nm (closely matched to the scintillation
light) is coupled to the PMT Ma a fibre-optic cable as shown in the figure. This is
used to stabilize the PMT gain as discussed in 94.3.4.
The DSSSD is mounted wvithin a plexiglass disk (as shown in figure 4.6) whose
thickness is only slightly larger than the 0.40 cm thick frame of the DSSSD (which
mounts the silicon wafer and houses the strip readout contacts). The low Z matecial and minimal thickness was chosen so as to reduce the probability of 0s scattering
off this mount. The plexiglass plate provides both a rigid muunt as weli as a well-
d e h e d overall DSSSD orientation; with the axes d e b e d in Figures 3.6 and 4.7, the
+A Panasonic digikey #P39û-ND light emitting diode.
4.1 DESIGN AND CONSTRUCTION 39
FIGURE 4.6: Schematic diagram of the DSSSD mounting showing the strip detec- tor's reference frame relative to that of the detection charnber.
DSSSD is rotated +(45.0 i 0.5)" with the y-strips facing the recoil detector. The
G10 frame represents the largest source of scattered Bs adding to the telescope's low energy tail; as the DSSSD is rnanufactured this way, we are not free to reduce this
contribution mechanicaliy, but we may suppress it in software by not including the
edge strips in the analysis (see 54.2). Coaxial wiring for the strip reaciouts is fed out
of the ,O-telescope's vacuum chamber through the 16-pin feedthroughs, and is input to
locally constructed [491 preamplifiers which are mounted directly on the flange. The
wiring frorn outside the shielded vacuum charnber to the preamplifiers is coaxial (and
additionally are shielded) to prevent noise €rom electrical pick-up.
Final GEANT Geometry
Once the ,O-telescope was built and optimized, a more realistic description of the
geometry vas input in the Monte CarIo simulation. A diagram of t he volumes included
in GEANT is given in Figure 4-7- The ,O window is at z = +6.601 cm and is defined
by the specifications [471 of the IF1 type berybumt. The bulk of DSSSD remains a
pure wafer of silicon, but now the geometry of the detector includes the 48 duminum O
readout strips (assumed to be 500 A thick) and, more importantly with regard to
t99.837% Be, with the biggest contaminations hom Fe (0.006%), Ni (0.02%), Ca (0.02%) and
Zn (0.01%). These percentages, and those hereafter, give the relative number of etements making up the medium.
4.1 DESIGN AND CONSTRUCTION 40
FIGURE 4.7: Geometry of TRINAT'S detection chamber that is input into further (detailed) Monte Car10 simulations. The volumes outside of the &telescopeYs vac- uum charnber are important because positrons cm (back)scatter off of tliem before entering the &telescope.
scattering effects, the G10t frame in which the wafer is rnounted. The plexigiasst
mount was simplified in the MC by apprauimating it as a ring of plastic wbose inner
diameter does not ovedap the GlO frarne. The four stainless steel rods connecting the
DSSSD mount to the (smaller) telescope Range are also included.
The scintillator has 6ve layers of Teflon* on the front face totalling 0.02 cm in
thickness, and many layers around the sides of the scintillator and light guide, totalling
about 0.7 cm in thickness. The BC408 plastic is defined by the hydrogen to carbon
ratio of 1.104, but also contains a 2% addition of PPOtt which is addeci to increase scintillation efficiency 1501. The light guide extends out past the Banges, but other
components behind the flanges (the clamp, the PMT, etc.) were not included because
they will not affect detection of the positron.
The rest of the geornetry inchdes the 2nd trapping charnber, the alurninum hoops
tdefined as 60% Sioz and 40% Kapton (C2HroN205). tdehed as CH2=C(CH3)COzCH3. 'defined as CF2=CF2.
ttdeûned as ClsH~lNtOr (2,5-diphenyloxazole).
for the electric field and the MCP assembly. The material of the MCP plates is
Corning 8161 glassi and the resistive anode is pure Si. A large number of 0's were
found to scatter off the MCP or one of its electrostatic components before entering
the ,O-telescope (see 54.5.1) and so it is important to keep these volumes in the Monte
Carlo, even though it requires w 10x the CPU time. The rods for the hoops, not
included in the simulations presented here, have since been included since they offer
a considerable volume of stainless steel that 0s can scat ter from.
The Double Sided Silicon Strip Detector
This section begins with a brief introduction to these semiconductor devices, but
for a more thorough review, the interesteci reader is referred to 1511. Sections 4.2.2-4 describe the energy calibration and resolution of the detector, followed by the position
decoding scheme in i4.2.5. The final section, $4.2.6, presents the results obtained from the correlation experiment data.
4.2.1 The Device
The AE component of the 0-telescope is a double sided silicon strip detector pur-
chased from Micron Semiconductor 1521. It is a silicon wafer of dimensions 2.4 cm x
2.4 cm x 491 pm upon which a thin (typically a Few pm) p+-type layer is deposited
onto the 'front' or ohmic side of the n-type silicon. Similarly an n'-type layer is
deposited on the grounded side ('rear') of the detector as depicted in Figure 4.8. This O
forms a pf n junction, each side upon which is evaporated a thin (typicaily 500 A) layer of aluminum; a reverse bias of -135 V is applied to the pf side (with the nr side
at ground) to ensure that the depletion layer extends throughout the Si wafer so that
al1 of the energy deposited by an ionizing particle gets coliected onto the duminun
readouts. The negative bias also increases the magnitude of the electric field across
the depletion layer which serves both to make the charge collection more complete as
well as to decrease the time needed for the electrons and holes to reach the readouts
+definecl as 8.8% Pb, 61.1% 0,24.6% Si, 4% K, 0.8% Rb, 0.3%Ba, 0.19%As, 0.08% Cs and 0.15% Na.
FIGURE 4.8: Schematic diagram of a p+n double-sided siiicon strip detector. Electron-hole pairs are created as an ionizing particle goes through the depletion region, and they are separatek collected on readouts on the pi (electrons) and n' (holes) side. Both of these readouts are divided into 24 strips, providing position sensitivity in both 2 and Y.
( i e . the pulse duration). To add position sensitivity ta this detector, the aluminum
readouts are both divided into 24 orthogonal strips whose width and spacing are nom-
inally 900 and 100 pm respectively. The lengths of the pf strips are aligned with I
and so, depending on which strip registers the hole collection, we obtain the position
of the hit in y. Similady, position information in ? is determined by the negative
charge collection on a given n f strip. With a depletion depth of d = 491 Pm, the
capacitance works out to CSkp = A/d = 4.4 pF, where A = 900 pm x 2.4 cm is the
area of a strip.
Upon entering (and exiting) the detector, a charged particle will interact with the
aluminum contacts, thus introducing two necessary (yet entirely negligible) dead lay-
ers. As the particle then goes through the bulk silicon, the energy deposited through
multiple Coulomb scattering in the depletion regiont creates electron-hole pairs; the
tthough fuiiy depleted, the wafer will have smd gaps in its depietion region near the surface of the areas between stnps - carriers generated here do not get coliected with the same efficiency.
positive holes are swept by the electric field to a readout on the ohmic side while
the negative electrons migrate to the readout at ground. Thus for each event, the
DSSSD fires two pulses - one for holes and the other for electrons - which are both
proportional to the energy deposited by the incident particle. The electrons and holes
are created in pairs so the two signals should in fact be identical, although in practice
differences occur due to electronic noise and, hopefully to a lesser extent, uncertainties
in the energy calibration of the strips.
The electronics for the DSSSD is illustrated by the sample strips depicted in the
figure of Appendix C (page 123). The timing of the y-strips is used to trigger events
while the energy signal (of both the x- and y-strips) goes directly to LeCroy 224912
chargôsensitive ADCs. The timing signals are fanned together in groups of four, each of which is then input to the 2228A TDC for separate timing; these six groups
are then fanned together giving any 'DSSSD event' either (a) used to provide an
event trigger (off-line calibrations) or (b) checked to see if it is coincident with the
scintillator's timing (on-line event trigger). The hardware thresholds for each of the
24 triggering strips were individually optimized and set to be above their respective pedestals (5 20 keV).
We will see how with a good energy calibration and an understanding of the detector's resolution, a condition on the difference between the two energy readings
provides a clean tag for /3 events.
4.2.2 Energy Calibration
Low-energy photons from the decay of 241.4m and 13%a were used to calibrate the
strip detector. '33Ba decays by electron capture [531 to a nurnber of excited states in
L33Cs. The resulting radiation consists of Cs X-rays a s well as several y rays, with the
photons of interest listed in Table 4.1. For calibration purposes, we use the weighted
average of the two unresolved M l y peaks and fit to one (slightly widened) peak at
80.898 keV. Similady, the Ka,- and K,,-sheU X-rays are averaged to 30.851 keV. The
cw decay of 241Am tu 237Np has a strong El branch (E, = 59.5412 keV [54J), providing us with a third calibration point.
The photoelectric effect and Compton scattering are the t m competing processes
for photon interactions over the energy range hv = 30 - 80 keV in silicon. In a
photoelectric process, a K-shell electron bound to the Si atom totaily absorbs the
incident energy; this 'photoelectron' is ejected 6 t h a kinetic energy Te = hv - Be, where Be is the sheli binding energy. As was the case for bariurn, the inner sheli
vacancy is Bled up, but in this case Auger processes are favoured and so the binding
energy is generally recovered. In Compton scatteringt the photon impacts only a part
of its energy to the struck 'Compton electron', the amount of which depends entirely
on the direction of the scattered photon. The result is an energy spectrum that
extends from zero up to a maximum (Te),, < hv corresponding to a backscattered
photon (see Equation (4.12)). For 30 keV X-rays, the photoelectric effect has the
largest cross-section so the DSSSD spectnim contains the large peak at E = hv with
a small low-energy tail arising from Compton coilisions. At 60 keV, these processes
have about q u a i cross sections and by 80 keV the Compton effect is an order of
magnitude larger; since the photoelectron's spectrum is concentrated over a srnall
energy range, the photoelectric peak is still evident in the y sources, although the
Compton tail cornplicates determining the centroid as precisely as with the X-rays.
In addition to those resulting from the photon sources, a large peak refimed to as
the 'pedestal' is evident in the energy spectnim (as in Figure 4.9); this peak occurs when the strip in question did not fire, but another strip triggered the event. For
the charge-integrating hDC1s used, the observed pulse-height of the non-triggered
strip corresponds to the ADC offset and, ideally, would be a 15 function. In ce-
tthis is covered in more detail in 84.3.3 with regard to the on-iine scintillator calibration.
TABLE 4.1: Low-energy photon sources used to provide a . initial calibration of the strip detector.
133Ba
2 4 1 ~ ~
Source
Cs Km, Cs Ka,
r ( W r (Ml)
r (El)
Rel. intensity
0.648 0.351
0.929 0.071
Energy [keVI
30.973 30.625
) 3û.851
80.9971(27) } 80.898 79.6139(6)
59.541
6.0 M) i(KI UO 140 160 Channel number (strip yJ
FIGURE 4.9: Sample fits to the DSSSD energy spectrum of the 7 source 2"h. The data is the coarsely binned histograrn and overiayed is the fit using a function consisting of a Gaussian (photoelectrons) plus a constant low-energy background (Compton electrons). The fitting range is indicated by the Iirics above and below the 59.541 keV pealc.
ality, the pedestai is a Gaussian peak centered at zero energy with a width that
directly reflects the noise in the electronics. The average width For the x-strips is
(O$") = 3.7 k 0.7 keV and that of the y strips (O:=) = 3.9 1 0 . 4 keV. The centroids
of each of the photoelectron peaks as well as the pedestal provide four points which
can be fit to a linear calibration relating the observed pulse height Mth the energy
deposited in the detector.
The Am and Ba sources were placed in front of the DSSSD within a light-tight
and electrically shielded volume: a thin stainless steel cylinder 040 cm x 60 cm (suf-
ficiently large to minimize scattering effects off the walls of the container) with a mount for the B-telescope assembly.
Each peak in the resulting energy spectra of the 48 strips was then individually
fit to a function consisting of a constant background that extends below the mean of
a Gaussian peak (the photopeak) representing the fuli energy reading of the ejected
photoelectron. Fits to the 60 keV peak from 24L,4m in strips xl and y, are given
in Figure 4.9. The pedestal is fit independently by gating on pulser events for the
scintiiiator, in which case ail of the DSSSD strips were read out with zero energy. The
widths, which are calculateci using the ha1 off-iine source calibration, are typical;
the pedestals contribute 3 - 4 keV to the noise while the pulsers are roughly equal
FIGURE 4.10: Energy calibration of the DSSSD strips x2 and y2 using the sources 133Ba and 241~m. The four points used in the fits (given above) consist of the sources listed in Table 4.2 plus the ADC offset. The calibrations have an offset with units of channels and the dope bas units of keV/channel.
$ = (0.890îO.OS)x[cbannal - (55.40î0.23)] \ = (0.741~0.000)x[channel - (97.49tO.E)I
to the photopeak widths of 5 - 6 keV. The fits to the barium peaks were done
in a similar fashion, but in that case a separate overdl constant background was
also required to account for the higher energy photons that Compton scatter in the
detector. The 81 keV photon interacts predominately through Compton scattering, so
the resulting photopeak in the observed spectrum is wveaker compared to that arising
from the Cs Ka-shell photons; these calibration points, therefore, have relatively large
uncertainties in their values.
Using the results of the fits to these spectra, energy calibrations were made using
linear regression on the four points. Figure 4.10 is the calibration for xa and y2
and numerical results of the off-line source calibrations are given in Table 4.2. The
calibration to the strips look reasonable but, as ive will see in the following section,
we can improve the calibrations using the on-line ,f3+ data themselves.
4.2.3 Extended Calibrations
In addition to the inherent resolution, the final energy reading from the strip detector
will dso be affected by systematic deviations in each strip's calibration in many ways.
As we wil l see in $4.2.5, event selection is based upon which and how many strips
pass a cahbrated energy threshold, as well as a m a - u m limit on AE = jE, - E,I,
O 50 XM 130 200 50 100 BO 200 ao Ch~nnd numbar of l k i p x, Channel numbar ot ¶trip y,
F xa/v = 0.81 CL = U J X
/'
4' '
7'. /"
,,.' ;./"
4 y r
t lm+
75 - - 5 f 5 0 - - 3 i 2 s - Y
O -
-25
m.
@ -
- 6 0 - % il
40 - $ g 20-
O
-20 -
xZ/v = 1.04 /' CL = S I X /" Y''
," /'
/' / ,' / ' -.
/ /
./'
. .
x calibration offset slope
(channels) (keV/chan) 58.llf 0.44 0.87kt0.028
55.40f0.23 O.89Of 0,015
43.49f0.27 O.893f 0.018
43.79f 0.18 O.885f 0.012
58.59k0.30 0.848f0.018
62.643~0.30 O.965f 0.023
60.03f0.24 0.982zt0.019
5l.98f 0.40 O.949f 0.030
65.10k0.20 0.816I0.011
65.99f 0.27 0.792f0.014
64.02f0.27 O.837f 0.015
67.49f 0.21 O.786f 0.011
64.04f0.21 O.925f 0.015
6l.74f 0.26 l.OO6f 0.021
6l.O8f 0.25 O.965f 0.019
190.32f0.39 O.685f 0.015
7l.85f 0.25 O.768f 0.012
61.15f0.23 O.8Ogf 0.013
91.98f0.26 O.82Of 0.014
63.54k0.28 0.828f0.016
72.66f0.31 O.8l5f 0.017
63.27f0.35 0.767f0.017
63.57f0.28 0.857&0.017
69.42f0.57 0.736f0.026
y calibration offset slope
(channels) (keV/chan) 77.15k0.24 O.85Of 0.017
97.49f 0.16 0.741zk0.009
75.26î0.12 0.822&0.008
72.62f 0.22 O.805f 0.014
74.98f 0.25 O.768f 0.015
52.98f 0.30 O.854f 0.022
66.82f 0.26 O.8l?f 0.018
75.lif 0.17 O.8l5f 0.011
l23.26f 0.14 O.7OOf 0.007
73.86f 0.28 O.773f 0.017 96.11î0.29 O.792f 0.018
92.87I0.20 O.Xi6f 0.011 42.4110.31 O.807f 0.020
53.46k0.25 O.847I 0.018
86.455t0.20 0.788I0.012
51.93k0.35 O.775f 0.021 72.52f0.28 0.759f0.016
108.56f0.18 O.76lf 0.010
126.26f0.21 O.725f 0.011
55.63k 0.24 O.843f 0.017
59.63f0.28 0.825f0.019
59.94f 0.29 O.8lOf 0.019
111.81f0.21 0.859I0.016
66.13k0.28 0.797k0.018
TABLE 4.2: Fit parameters of the DSSSD strips off-line source calibration. The caiibration is fit to a function of the form EkeV] = slope x (channel- offset).
and so poor energy caiibrations will cut legitimate events. in order to minimize the
differences in each strip's energy reading, we take advantage of the fact that the two
signals in i and $ should be identical, and effectively re-align the strips by forcing
that fact to be true on average. To that end we can use the on-line 38mK data which
is advantageous because they provide a continuous spectrum well beyond the 81 keV
barium peak; the ce-alignment of the calibrations using the data themselves is, by
definition, over the entire energy range of interest. In applying the off-line calibrations to the on-line data, corrections for possible
DC offset aud gain variations in the electronics were made by comparing the peak
positions of the pedestals and pulsers. Throughout the experimental mns, the offsets
of the original calibrations are adjusted by the difference in the pedestal centroids
(over = 5 minute intervals) from the centroids observed when calibrated. Sirnilarly,
the slope of the calibrations was adjusted by the ratio of the pulser centroids. If si, is the pedestal and x",,, the pulser centroid wheu the off-line calibration was done,
then the corrections applied at time t are:
offset'(t) = offset - (xpd(t) - x b ) (4-1)
The calibrations were done only a few days before the April 1999 experiment and
since the DSSSD is an inlierently stable device, these corrections end up being very
small, typically on the order of 0.1%.
After applying these corrections to the 38mK of data, for each x-strip that passed
a low-energy threshold of 40 keV with only one corresponding strip in y, the energies
E, and E, were plotted as in Figure 4.11. Assuming that variations in the y-strips'
calibrations average out, the deviation from (E,) = Ki provides us with a correction
with which to align each of the x-strips to the average of y. In making the fits,
Poissonian statistics needed to be used in order to be able to include the scarce events
at higher energies. Note, however, the extended low-energy tail especially evident in
the edge strip xl (the inner stcips, e.g. x2, are much cleaner indicating edge effects in
the DSSSD). Eue y event in Figure 4.11 is plotted, so this actually only represents
< 10% (' 1%) of the total number of counts in the edge (inner) strips. It is still an N
important concern, however, because this low-energy tail has a stronger weighting in the fits if al1 the data are used, thereby biasing the calibrations to axi < 1 and PXi > 0. In order to minimize this effect, yet still retain the high-energy events, an iterative
method was used: in the first step we assume aXi = 1, ,& = O and fit Equation (4.3) only to events where Ex - 3OkeV 5 Ey _< E, + 3OkeV. Based on the new values of
aXi and a,, we again fit using events where the new [Ep - EkIs 30 keV. This process was repeated until the changes in the parameter values were less than their respective
uncertainties. Table 4.3 lists in detail the corrections to each of the x-strips using this
alignment scheme.
The iterative method used to fit the extended calibrations ends up applying rather
large corrections to the *"'Am and '33Ba calibrations. Generally, the slopes are de-
creased and the offsets increased, but by far more than can be explained by the bias
from the low-energy tail as described above. Most surprishg is the change in the
offset; the pedestal peak is well-defined and if it really does represent zero-energy, one
would expect the ai parameters to be very close to zero. Random changes in the a,,
would be expected (versus a systematic difference) since the larger energy range of
the on-line data is more sensitive to the dope of the caiibrations. The extended cali- brations and the calibration of the DSSSD as a whole are therefore not without their
problems, and so the uncertainties quoted by the fit are certainly underestimated.
The extended calibrations are effective in aligning al1 of the x-strips' calibrations
together, which is necessary in order to be able to compare the energies of each equally.
In order to sirnilady place the y-strips on equal footing, we do the same procedure
to re-fit them to the average of the corrected x-strips. The corrections are listed in
Table 4.4 and Figure 4.12 shows the fits to y, and yz. The variance of the points
around the fit is reduced compared to Figure 4.11 because the x-strips have already
been calibrated to a cornmon energy auis. The fit parameters stiU show the same
trends as in the x-strips (& > O and ayi c l), but are not as large or systematic.
The high-energy tails, however, can be seen to still have a bias since they do not
foiiow the fit line to within the quoted uncertainties, again indicating that they are
TABLE 4.3: DSSSD caiibration fits of = (4) = o, x Er, + ai. These fits align the calibrations of the x-strips to the average calibrations of the y-strips. - s trip
- strip -
13
14 15
16
17 18 19 20 2 1
22 23 24 -
F~GURE 4.11: Cornparison of Ex and Er energy readings using the calibrations determineci using off-line sources for strip xi (Mt) and q (right) . The solici iine is the fit Eg = a x Ex + a and the t l i ~ ~ h t ~ l lines contain the fitting region defineci by JEp - E:j_( 30 keV,
underestimated.
As with the x edge strips the y-strips contain a background, but in this case it is
manifest as a high-energy tail. The vast majority of t hese events have too much energy
in jj and correspondingly too little energy in 5. Yorkston, et. al [551 have discussed
such effects in the case where the incident particle was a heavy ion, and they found
that inter-strip events gave cise to pulses of the wrong polarity €rom induced charges.
They performed similar tests wit h 207Bi but did not observe this effect in this case; the
tails seen in our edge strips are not at present understood, but the effect is relatively
small and the events can be excluded in software if desired (see also $4.4.3).
4.2.4 Characterization of the Resolution
The dominant noise apparent in the DSSSD spectra is electrical, but additional com-
ponents mise Erom fluctuations in particle-hole generation and charge collection. As
mentioned eadier, the electrical noise determines the width of the pedestal wbile the
overail inherent tesolution of the detector should be reflected by the width of the
photopeaks. The average widt hs of the 133Ba M 1 -y peaks are 5.6 f 0.5 keV for y and
5.5 f 0.9 keV for x, in agreement with the resolution of these devices as indicated in
the literature [511.
The final resolution of each strip ivas determined by comparing the difference of
the corrected energy reading of a given strip to that of the orthogonal's corrected
energy. This is basically the same procedure as in the extended calibrations, except
this time we just look at the deviations in order to generate an average resolution
hnction. -4s can be seen for x16 in Figure 4.13, the final extended calibration of this
strip versus al1 of those in y still suffers from a slight bias, as is especially evident
from events in the high-energy tail of the peak (above x 400 keV). The dashed lines
above and below EL = Ek are at st24 keV, representing a AE 5 *3gX,, acceptance
for this stcip (see below). The low-energy thresholds are set to 40 keV, well above
the pedestals of the strips. This strip was chosen as the example because it has the
largest offset and smailest energy range as is evident by the saturated ADC readings
at Ex,, x 725 keV. In order to ensure that no biasing of the on-line data occurs due to
such saturated energy readings, a high-energy threshold of at least 725 keV needs to
TABLE 4.4: DSSSD calibration fits of Eii = (Ex) = ayi x hi +hi. These fits align the calibrations of the y-stries to the average calibrations of the (already corrected) x-strips.
strip
1
2 3
4 5 6
7 8 9 10 11 12
strip
13
14 15
16
17 18 19 20 2 1 22 23
24
L
a,, = O.SZ6r0.0006 p,, = 7.0Zt0.11 keV - xZ/v = 0.55 -/
F~CURE 4.12: Cornparison of the energy readings of stries y1 (left) and y2 (right) to that in x. The result of the fit yields EP = ayi EJri +& which aligns each y-strip to the average of the (already corrected) x-stries-
-mo -1 I
O 200 400 600 600 Energy in X, [keV]
FIGURE 4.13: Plot of E:,, vs. ( E i ) - E:,, with the proposed thresholds for the on-line data indicated by the claslieil lines. Strip xi6 has the largest ADC offset and hence smailest energy range; the upper energy threshold needs be placed at 725 keV so as to ensure that no events are off the energy scaie of this the most iimited strip. The solid lines at f 24 keV represent an 2-4 energy agreement condition used to discrimate against noisy events.
be imposed; the threshold may be set lower to value of 400 keV so that the bias from
the irnperfect alignment of the calibrations is contained wvithin the stcips' resolution. The resolution functions defined by:
RZi = (E;, - E;,) and hi = (E;, - EL,)
were generated and a sample of them are depicted in Figure 4.14. The fact that the
Gaussian fits agree reasonably well over the peaks of the distributions, and that they
are nearly centered around zero, is very encouraging. This indicates that the resolution
of the strips is dominatecl by random processes and that the strip calibrations are
consistent. The widths of these Gaussian fits, listed in Table 4.5), average 8.1 keV.
The (relatively) large tails of the edge strips, however, show that they suifer bom
additionai systernatic effects. Since these tails are not a large percentage of the events,
the edge strips can be included in a finai analysis, but one might wish to consider
excluding these events if statistics is not a concern.
atrip: X, 30 mantr: 132579 (795% ot are. total)
L
rnean - 0.OSt0.03 1
mean = Ofllt0.06 - - w- = 8.29a.06
FIGURE 4.14: Overaii resolution Functions for strips xi, xi:! (left), yi and y[* (right). The resolution is calnùated as &, = (EL, - E;,) and Ri, = (Eki - E L ) . The daçhetl Lines are at f 3u of the Gaussian fit (solid Line) , which fit weii over the peak of the distributions; the xi and y1 have especiaiiy large tails arising from edge effects.
- - Strip: Y. Su nanta: (59255 (W.S% of are. total)
r r i
m
na 4
9-- O Ci
M p : K, Su wmts: U22lD (ï9.SX ot ave. totd)
-50 O 50 la -100 -50
Strip: X, rnean - -0.0610.03 JO aiientc US700 0,- = 7.8720.03 (85.8% of are. total)
mean = -0.22*0.06 u , , ~ = 8.5h0.06
O4000- i
U P
I
TABLE 4.5: Final widths of the DSSSD energy readings by comparing x and y energy readings. The average widths are (nJ = 8.07 & 0.18 and (ay) = 8.08 f
- strip -
1
2
3
4
5
6
7
8
9
10
11
12 -
x-st rips
8.29f 0.06
8.14k0.04
8.10f 0.04
8.l4f 0.04
8.Oïf 0.03
8.O4f 0.03
7.96f 0.03
8.OOf 0.03
7.97f 0.03
8.O3f 0.03
7.89f0.03
7.87f0.03
- strip x-strips
7.76f 0.03
7.80f 0.03
7.94f 0.03
8.18f0.03
8.18f 0.03
8.O7f 0.03
8.17f 0.03
8.24f 0.03
8.2lf 0.04
8.35f0.04
8.36i 0.04
8.96f 0.07
Calibrate each strip and impose energy threshold conditions
1 If strips that fired are adjacent, add up their energies to form
1s there only one hit in I "O- a total 'group energy.' Now, and only one hit in .y? is there only one p u p in 5 and
one group in Y?
Veto eveut (miiltiplc Kt)
Accep t evei~t Veto event ( t i t i kriotvri energ)
FIGURE 4.15: Block diagram of the DSSSD analysis scheme. Essentidy, we sirnply require that events in the strip detector are consistent with a single hit, and that the energy deposited in the two sides (2 and y) agree to within the resolution of the strips.
4.2.5 Position Decoding and Analysis Scheme
The analysis scheme developed For the DSSSD relies on two important premises: (i) a
'good' event is one where only one pixel, a single hit in 2i and one in y j , fired and (a) the energy reading in Z is consistent with the energy reading in 6. The first ensures
that one particle went through the detector only once (in particular discriminating
against ps that backscattered out of the scintillator), while the second is a necessary
condition for a final energy reading. A surnmary of this analysis scheme is given in
Figure 4.15.
The fact that the strips are calibrateci allows us to impose energy threshold con-
ditions that applies equally to al1 strips. The first of these conditions is a low-energy
'strip thresholdl that is placed well above the hardware (triggering) thresholds, and
any strip whose energy is greater than this is considered to have fired. Ideally (indeed
most likely), only one strip in 5 and one strip in 9 passes this condition giving us
an event at xp = (i - 12.5) mm, y,, = ( j - 12.5) mm where i, j represent the strip
number of the x- and y-strips that ked. We would like to make this threshold as low
as possible, but we must make sure that it is high enough that we don't have strips
firing due to a random fluctuation above their respective pedestals; if this happens we
wili veto legitimate events as (phantom) multiple hits. With condition (i) satisfied
and the position of the hit knom, we now need to determine the energy the particte
deposited using the two energy readings. As mentioned earlier, the energy in Z should
be the same as $ to within the resalution of the individual strips, so we check to make
sure that they pass an energy agreement condition:
where
and ucUc is an adjustable (user) parameter that determines how tight this energy
agreement must be. The dEIi, dEyi are the la widths of the resolution functions,
as given in Table 4.5. If this condition is tme, then we take the final DSSSD energy
reading to be the average of the two calculated energies weighted by their respective widths:
Events that do not satisfy ( 2 ) are not immediately vetoed because we expect that
,8s will occasionally be scattered into an adjacent strip. For this reason if two strips
fired in, Say x, instead of just one, we check to see if they were xi and xi*,. If so, we
add up the energies deposited in each and weight the position of the hit according to the energy deposited in each, i-e.
and
In this way we define 'group' hits which are treated the same as single strip events:
if there was one group in x as well as one in y, we then check for energy agreement
between the group energy readings and accept the event if it passes.
The most sensitive test of the setting of the strip threshold is again through use
of the on-lie 38mK data themselves; the x- and y-strip positions, combined with
the relative number of single-pixel versus multiple strip events, indicates good noise
discrimination without a loss of good events. Figure 4.16 is a plot of the XDSSSD
position for events in which there was one (group) hit in 5 and a corresponding one in
9 which passed the AEDsssD 5 3uDsssD energy agreement condition. The large peaks of the individual strips (i.e. where no adjacent strip passed the strip threshold) reflect
the consistency of the x- and y-strip calibrations. A poor alignment causes the energy
condition (Equation (4.5)) to veto more events, the number of which would depend
on how rnisaligned a given strip's calibration is compared to the average of the other
side's; this is seen in the position spectrum when the initial off-line calibrations were
used (dashed line, slightly offset for clarity). The extended calibrations can be seen
to remedy this situation by making these peaks much more homogeneous. Almost 7% of the events that would otherwise be vetoed are retrieved once the strip calibrations
are ptoperly aligued and, more importantly, the efficiency of the DSSSD is rnuch more uniform across the strips.
The distribution of the inter-strip events is less sensitive to the relative consistency
of the calibrations; the shape of the distributions for the off-line and extended cali-
brations is very similar as seen in Figure 4.16. These events are, however, sensitive to
the applied strip threshold; if set too low, a strip adjacent to where the P left (au of its) charge is more Iikely to fire from random fluctuations above threshold. This will bias the shape of the inter-strip events towards having too many events just below
the main single-stcip peak (there is not a corresponding bias just above because of
the choice of binning in this figure accentuates the very low Ei-l, while the corre-
sponding Ei+l is absorbed in the main peak). Alternatively, if the threshold is too
high, we would also see an overali l o s of events because the cases where only a small fraction of the charge went into the adjacent strip would be less likely to p a s . For
example, if 100 keV is deposited in xi and 40 keV in ~ + l (with the fuil 140 keV in
y j ) , a threshold of > 40 keV would not add back the ~ + l energy back and may veto
extended - - - OU-iine 1
- 10 -5 O 5 10 Position in x [mm]
FIGURE 4.16: Position of DSSSD P hits in 5 fiom the on-line 38mK data. The claslied line corresponds to the initiai photon calibration while the solid uses the extended caiibration. Note the improved hornogeneity in the nurnber of counts of the individual strips (linear scale, above) as well as the symmetry of the inter- strip hits (logarithmic scale, below). The corresponding spectra for ?j shows sirnilar characteristics.
the event since Ex, # E,,.. Thus, too high a threshold suppresses events near (i f 6) where b c 0.5, but wvill not seriously affect events near the midpoint since these leave
nearly equal amounts in each. Tbe shape of the inter-strip events (averaged over al1
x-strips) are plotted in Figure 4.17 and shows how the data are affected by the value
of the low-energy strip threshold. The lits, which are to a quadratic function, are
aU in agreement with the data eircept for the hst point; the 20 keV threshold is too
low because it allows too many events here while the 65 keV one is clearly too high.
Between these two thresholds, the shape and even the overall number of counts is seen
to be relatively insensitive to the value used. The large 65 keV threshold is clearly ex-
cluding events where a small (but real) amount of charge was deposited in an adjacent
strip. By looking at plots similar to Figure 4.16, but with h e r steps in the low-energy
threshold, it was found that the inter-strip distributions were constant over the range
35 - 55 keV. The results containecl nrithin this thesis ail use a low-energy threshold
of 40 keV.
20 keV threshold 35 keV threshold
a Li
A 50 keV threshold O 65 keV threshold
i 1 I I I
0.0 0.2 0.4 0.6 0.8 1.0 Inter-strip position [mm]
FIGURE 4.17: .Average position of inter-strip DSSSD events using different low- energy thresholds. When the threshold is too low (20 keV) , we see too many events below the single-strip peak and if set too high (65 keV), events to either side of the main peaks are reduced. The lines are quadratic fits to al1 but the lut point.
As mentioned in $4.2.1, a charged particle such as a relativistic positron loses energy
through multiple Coulomb scattering, which is an inherently random process. The
energy deposited by the positron will depend on the angle(s) it is scattered into and
how long the random walk through the detector is; the energy density distribution
follows a Vavilov distribution 1561 which contains a large peak corresponding to parti-
cles which suffer only small angle scattering, as well as a high-energy tail for the ones whose random waik through the detector is relatively large.
The on-line 38mK data in Figure 4.18 show the final DSSSD energy reading as
analyzed according to the scheme outlined in 84.2.5, and are compared to a GEXNT \
simulation. The Monte Car10 has been renormalized so that the peaks of the two
spectra are equal. Both the simulation and the data have the following conditions
imposed:
a 40 keV low-energy DSSSD threshold (from the previous section).
- MmK data
GEANT MC
moment 1 data MC
mean 175.2 164.9 st. dev 53.2 48.5
skewness 2.02 2.35
FIGURE 4.18: DSSSD energy spectrum of the on-line 38mK (solid Line, al1 stries) and the comparison to GEANT simulations (llii~ti('tL). The first three moments of both distributions are given to characterize the ciifferences.
a 500 keV high-energy DSSSD cutoff (to reduce b i s as noted for Figure 4.13).
an Ex, - Eyj energy agreement condition using rCut = 3.
a 2500 keV low-energy scintillator thresholdt
There is a clear discrepancy between the experimental data and GEANT with a
> 10 keV difference in their average energy losses; the table in the figure lists the first,
second and third moments of the two distributions. The sharp rise on the low-energy
side of the peaks agrees (up to a constant shift) and this shows that the simulated
data have had the detector resolution properly incorporated. The high energy side of
the peak is clearly sharper in GEANT, and this may be indicative of an underestimate
in the nurnber (or angle) of scatten within the DSSSD. Part of the problem rnay arise fiom the extended cdibration corrections (of §4.2.3), and simply be a probiem with
tthis condition on the E detector's energy reading eluninates the 38K ground state 7 contami- nation, which is not indudeci at this stage of the analysis (see 84.3.5 and beyond).
the overall calibration. Spectra of the individual strips are al1 aligned by virtue of
the resolution functions being centered about zero, and therefore the data are not
wider due to strip calibration variations. If the difference were solely due to the
overall calibration, then it could be corrected by fitting the data to the MC, with the
energy scale free to vary. Attempts to do this did not fully correct the situation as
the high-energy side of the peak remained underestirnateci, so it is more than just the
calibrations.
At this time, we do not fully understand this difference, but we do know that
in relation to the telescope's total E i- A E energy reading, this discrepancy is very small: at 1 MeV, the average difference is only 1% of the total and it is only 0.2% at
5 MeV. The impact on simulations of the scintillator's spectrum, which must include
the energy lost in DSSSD, is discussed in 54.4.1, and 84.3.5 shows that the total energy
spectrum is well reproduced if the average difference is considered. It is therefore not
crucial that we resolve the discrepancies found within this section, but improvernents
can be made in the understanding of this device if more careful calibrations are made.
The full energy peak of a conversion electron source, such as 207Bi, would help to provide a cleaner initial calibration which would make the extended calibration easier
and presurnably involve smaller corrections. The source must, howvever, be open
(unlike the one used throughout this thesis) so that the electrons do not suffer (much) energy straggling before entering the DSSSD.
4.3 The Plastic Scintillator
This part of the p-telescope is important to TRINAT'S B decay experiments because
it provides the (rnajority of the) energy measurement of the emitted positron and the
time of the event relative to the recoil detector. It has been my hope from the start
of my studies to design and optimize the scintillator weil enough that the energy and
timing resolutions do not limit our measurement of the correlation parameter. The
timing of the scintillator wiii be shown to be good enough that our recoil time-of-
mght measurement is b i t e d by the trap size. Preliminary tests [5fl of our sensitivity
to uncertainties in the calibration of the gain indicate that Au, = 0.1% for a 0.1%
change in the slope of the gain. As anyone who has worked with ,Bs wiil appreciate,
this represented a formidable task.
4.3.1 Optimization
The block diagram of Figure C.l in Appendix C (page 123) outlines the scintillator's
electronics. The energy signal is taken from the dynode output of the PMT and
inverted through a LeCroy 428F linear fan in/out. The anode provides the timing signal to the Tennelec 455 constant fraction discrirninator which is used to set the
scintillator's hardware threshold to approximately 300 keV. The stabilization unit,
discussed in greater detail later (see §4.3.4), minimizes gain variations by monitoring
a LED's pulse-height in the PMT. Once the scintillator and its electronics were assembled, tests were performed to
determine the wrapping that would best collect light in our geometry. The wcapping
of the scintillator can be broken up into two areas: the front face of the plastic,
and the wrapping around the side of the scintillator and light guide. For the front face, extra care was taken to minimize the energy straggling of positrons as they
enter the scintillator. To that end, 207Bi spectra obtained using a Si detector were
compared when sheets of different wrappings were placed in front it. 207Bi decays via orbital electron capture [581 giving two mono-energetic internal conversion electrons
at 975.6 and 1048.1 keV, which are resolvable with the Si detector. Comparison of the
energy loss effects of different wrappings and thicknesses could then be made using the
peak-to-valley ratio of these two peaks. This was done for aluminized mylar (Al on
C5H4O2), Teflon (CF2=CF2), white paper and Tetlart, the results of which indicated
that Teflon [59j caused the least amount of energy loss and straggling.
Figure 4.19 shows the spectra of a 207Bi source using a few of the wrapping schemes
considered. Once again, the diffuse reflector Teflon was found to be best suited for
our geometry because it was with this wrapping that we observed the lacgest signals.
One can see that Teflon around the sides clearly yields better overall light collection
compared to Ai mylar. With Al mylar only on the fiont face, the same overall gain
is observed as the Teflon (once enough layers was applied), but one can also see the
+This is nomally used to keep scintillators light-tight, but these tests indicated that it adds to the energy straggling of the incident particle and so was not used,
- A l mylar Al mylar % .
- 1' ! - - - - Teflon (xl) Tellon l Al mylar Tellon l .!, I La .: 1 . . . . - . . - Teflon (x4) Tel lon m .
! . , .: d 10' : f
T 1
200 400 600 800 1000 Channel number
F m RE 4.19: 207Bi spectra using different scintillator wrapping schemes. The Teflon wrapping around the sides of the scintillator and Iight guide c m be seen to provide the greatest light collection; aluminized mylar on the front face yields the same light collection efficiency, but Tefion minimizes the energy straggling of the electrons as they enter the scintiiiator.
energy straggling effects in the electron conversion peak. For this reason, the final
wrapping of the plastic and light guide wvas done entirely with Teflon (as specified in
GEANT'S geometry on page 40).
4.3.2 Timing
Good timing resolution between the recoil and P detectors is essential for accurately
determining the recoil momentum from its time-of-flight. To understand the tim-
ing characteristics of the scintillator, tests were perforrned !vit h anot her scintillator
(04 cm x 4 cm) using a 22Na source. When placed between the two scintillators,
the back-to-back annihilation radiation from the P+ decay of this source allows a
measure of the relative timing between the two detectors. The resulting timing peak
had a FVmI = 0.9 ns which, assuming both scintillators have equal timing, gives a
a,, x 0.4 ns.
It is the relative timing of the scintiiiator with that of the recoil detector that
is relevant to the experiment and, since the MCP does have a small efficiency for
4 I I
a,, = 1.00*0.09 ns
FIGURE 4.20: Relative timing between the P-telescope and the recoil detector using an source (left) and the on-line data (right). Both fits yield a scintillator-MCP timing resolution of ust 1 ns.
detecting ys, tests using a source were made using the essentially coincident
emission of the 898 and 1836 keV lines. An 8 ns delay cable was added to the
scintillator's timing half-way through the test in order to allow an easy channel-to-
time calibration. The results of a double-Gaussian fit to the two peaks are depicted
in Figure 4.20 (left) and give a relative scintillator-MCP timing, a&, of about 1 ns.
The timing from scintillator-MCP coincidences in the on-line data contains, in
addition to the Ar recoils, a peak at zero TOF. This 'prompt peak' (right of Fig-
ure 4.20) is thought to arise €rom events where one of the annihilation ys hits the
recoil detector; if an electron is ejected in one of the channels, the MCP will register
it as a hit. The prompt peak's energy spectnim in the scintillator is badly distorted towards lower energy, indicating that the other annihilation quantum is what pre-
dominantly fired the 0-telescopet. The events above 511 keV probably correspond to
the positron backscattering off the MCP and then into the telescope (the transit time for a relativistic B is 0.35 ns).
The tail at longer TOF (103410 ns) may corne Çom events where the positron
+A DSSSD coincidence is required in thh timing spectnun.
fires the P-telescope and this tirne a low-energy photon produced by the decay fires
the MCP. The daughter Aro atom has a good probability of being left in an excited
state (11 eV) after the 0 decay; these excited states decay to the ground state by the
emission of a W photon.
The fit to the prompt peak is a Gaussian plus two exponentid tails with decay
times of 2 and 8 ns (for the atomic excited states), d l with free normalizations. This
fit agrees with the '*Y measurement and so we take our detector TOF resolution to be O T ~ F = 1 as.
In the @ - v experiment, the resolution of the deduced energy of the recoil is
dependent not only on the relative timing of the detectors, but also on where the
decay occurred. The size of the trap affects the uncertainty in the position of the
decay (which is assumed to be the trap center) and hence the distance the recoiling
atom transverses before firing the MCP. With the typical trap dimension of 0.5 mm along the detector auis and considering the fastest recoils (Ar+4*+5~--.), this efTect adds
a f 4 ns 'time jitter' which is greater than the detectors' relative timing resolution.
The timing characteristics of the P-telescope are therefore adequate for the energy measurement of the recoil by TOF. If the trap size can be reàuced by a factor of four,
however, the uncertainty in the scintillator's timing will become significant. -4nother
reason for trying to improve the timing resolution is to use the 38mK measurement
to deduce the trap size; this could then be compared to that obtained using a CCD camera, which is not without its own systematic errors.
4.3.3 Energy Calibrat ion and Resolut ion
The largest systematic uncertainty in the @ - u experiment introduced by the B- telescope is the energy calibration of the plastic scintillator, and therefore considerable
effort was directed toward understanding it as completely as possible. Unfortunately,
a well-characterized source of mono-energetic positrons between 1 and 5 MeV is not
readily available at TRIUMF, so the calibration tvas done by fitting the Compton
edges of various y sources.
The analysis of the Compton spectra was performed using two schemes: fitting
the data analyticdly to the Klein-Nishina formula, as well as fitting to a Monte
0 .O O .S 1 .O 1.5 2.0 Compton electron energy [MeV]
FIGURE 4.21: Kinematics of the Compton effect aad the energy spectrum of the Compton scattered electron for an source (ET = 0,898 and 1.836 MeV) and a an annihilation quanta (ET = 511 MeV). The solicl line is the Klein-Sishina calculation and the dashed lines represent GEANT simulations.
Car10 simulation. Agreement between the tnr, methods served as a consistency check,
but the detailed analysis was limited to the Monte Carlo fit sirice the analytical formula does not indude multiple scattering, surface effects, 6 rays, and otber effects
contributing to the low-energy tail below the Compton edge.
The energy calibration is assumed to be linear (see 8.4.1) so that the observeci
pulse height, X ~ C , c m be caiibrated using:
Here xo is the ADC offset and X has units O€ channels/keV.
The dominant component to the spectrum of 7's in plastic is the Compton effect
( c . photoabsorption in a Ge(Li) or Nd scintillator) which is the inelastic scattering
off an atornic electron that is assumed to be free, as depicteci in Figure 4.21, The
energy of the Compton electron foLlows from energy-momentum conservation and
depends only on the direction of the scattered photon:
7i-aey2(l - COS 0) = l + ? ( l -ms0)
with y = E,/m,. The photon w i l transfer the greatest energy when it backscatters
(cos0 = -1) so that the maximum energy in the Compton spectrum occurs at the
so-called 'Compton edge': -
To calculate the differential energy spectrum of the Compton electrons, we turn
to the quantum mechanical calculation of Klein and Nishina [601. Letting s = T,/E,, the result is:
where r, = e2/4neome = 2.818 fm is the classical electron radius. GEXNT uses
Equation (4.13) to simulate Compton scattering, and one can see the comparison
of the Klein-Nishina formula with simulations in Figure 4.21. The extra bump at
energies above the Compton edge anse from events where the y bas scattered more
than once within the scintillator. A larger low energy tail is evident due to particles
escaping before depositing their full energy as well as from electrons that Compton
scattered from various neighboring volumes before entering the scintillator. The effects
of bremsstrahlung in the low Z scintillator at these energies is small (see Figure A3),
but is calculated in the Monte Carlo.
The final fitting function for the y calibrations needs to account for the resolution
of the scintillator whidi depends on the number of generated scintillation photons,
the number of these collected on the photocathode, noise in the electronics and in-
homogeneities in the plastic. If we assume that the width is dominated by photon
statistics and is therefore Gaussian, we use a function of the forrn [611:
where p and uo have units of keV. The noise introduced by the electronics, no, is
reflected in the peak observed when the scintillator does not fire (the pedestal).
The distribution used to fit the 7 calibrations is the convolution of the Monte
Car10 simulation with the resolution, Equation (4.14), Le.:
The parameters free to Vary in the fit to data are the energy calibration (x,, A) and
the width of the resolution (a,, p ) , The overall normalization is h e d such that the
data and simulation have an equal number of counts over the fitting range. Fits to
88Y are given in Figure 4.22 and are seen to agree well when the fitting range is
limited to just the Compton edge itself (top and middle; the arrows denote the low-
and high-energy Iimits of the fitting range). The events above the Compton edge
arise from cases where the 7 Compton scattered more than once in the plastic, and
this effect is generally reproduced in the MC; for the low-energy 7s (511 of 22Na and
137Cs), the MC does not do as good a job. The fit to the whole spectrum (bottom)
is not reproduced as well, and the resolution is forced to increase to help compensate
for this fact. Note, however, that when both edges are included, the offset and slope
can be simultaneously fit and that the fit values for the calibration do not change
much. The correlations between these parameters (as well as a, and p) , however, are
as large as 90%; the results can depend strongly on the initial guesses for their values
and so the results are not reliable. We must therefore fk the offsets xo and o, when
fitting the spectra and should only fit over the Compton edge itself where the MC reproduces the data well. As an initiai guide, the parameters xo and a, are taken to
be the centroid and width of the pedestal.
The channel nurnber, xmc, and width, asin, at each (Te),, are calculated using
the fits to X and p from the y sources: 22Na, 88Y and I3'ICs. These six points+
are then used to fit Equations (4.10) and (4.14) which then give the first iteration of
the calibrations: dl A', d, and p'. In this case, the offsets, xQ and a,, are free to vary and so are fit, even though in fitting the actuai y spectrum they are always fùred.
The y spectra are then re-fit with these new values of the offsets fixed, and the whole
calibration process is repeated. This iterative method is continued until changes in
the fit values become negligible.
tThe two lines in '%O are not resoIved by the scintillator and so only constitute one fitting point.
1400 1 L
x, - 43s channels
1200- A = 294.23IO.41 ch/MeV 3 E p = l.69IO.W keV 0 1000' x'/v = 1.1a3 CL. = 20% O ~i 800 - O
600 - al O E 400 - 1 2 200 - I b
L
x, r 43.50 channels
m 1200 A - 295.8310.32 &/MeV 4
p = 2.0210.14 kaV 2 1000 - l -- x2/v = 1.1002 C L . = 21%
O c 800 O I 115LI keV cl 600 - ' 400 Ei I 1 2 200 - I
1
O -* )r
FIGURE 4.22: Fits of a 88Y Compton spectnun to a GEANT simulation. The fits are to the 898 keV 7 (top), the 1836 keV 7 (middle) and the whole spectrum (bottom). The solid line is the MC over the fitting range (indicated by the arrows) and the dashed shows the rest of the hilC spectrum.
O Q I O min rn 22 2 23 2 V O in rD 9k 9 k * F 3'pc c3 m o * 010 in * C i 1 a) -im m m i n ri
n m O O* MS eu I.@? 0 @!O 4 Nrc eu 4 4
Nin rn ma) rc ?* e? @?Y m 3 0 O O 0 O
N O e3 u 3 Q ' i n q'? -! k'3. F9 nt- ri m o * z g m 3% 2
Table 4.6 lists the results of such fits for two iterations of a calibration for the
Aprill999 experimental runt The fit to the calculated X A D ~ of the 2" iteration and
the residual plot are given Figure 4.23. If we are interested in calculating the energy
for a given channel number, the calibration in the figure is expressed as:
The solid lines in the residual plat are the uncertainties in the calibration as given by
the above expression. The discrepant point at (Te),, = 1.94 MeV is from the Comp
ton edge of the 38K 2.2 MeV y. This Compton edge could not be fit well because of the
large 38mK 'background' which needed to be included in the fit; as will be mentioned
later, there appears to be a 2% discrepancy in the scintillator response between ys and
/3s and so simultaneously fitting the B spectnun will necessarily bias the Compton
edge fit. This point was therefore not included in the fit to the calibration.
Over the limited range that the y sources cover, and neglecting the poor 38K result,
the scintillator is found to be linear and calibrated to within f4 keV. Indeed, since
the experimental region of interest extends out to 5 MeV, ive still must be concerned
about the linearity at higher energies. G 24L.4m/Be source is normally used for the
neutrons it produces via the gBe(a, n)I2C reaction, but attempts were made to use
it as a 4.44 MeV y source from when carbon atoms left in the excitecl state decay
back to the ground state. The neutron background, much like with the 38K, made the
fit to the Compton edge unreliable. Still, to give some understanding as to possible
non-linearities in the scintillator gain, Figure 4.24 is a plot of a fit to the Compton
edges of the lower-energy ys as weli as this 241,b/Be source. The solid line represents
a linear fit through ail the points while the dashed line is a fit with the non-linear
calibration given in the figure. The diierence between the two types of calibration is
seen to be less than f 25 keV which should be a conservative upper limit considering
the poor fit to the 241,4m/Be source. Clearly, it would be desirable to add points
between the 88Y (1.9 MeV) and 24LAm/Be Compton edges to better determine the
linearity between 2-4 MeV.
Returning now to the results of Table 4.6, the resolution was fit to the Uxin of
the 2nd iteration with the results given in Figure 4.25(a). The 38K and 241Am/Be Compton edges are plotted, but not included in the fit (the 241,4m/Be was taken from
x,, = (43.010.3 ahan) + (296.210.4 ch/MeV) x T,
O 200 400 600 8ûû 0.0 0.5 1.0 1.5 2.0 2.5 Channal number (TJ,, [MeVI
FIGURE 4.23: Calibration of May 2nd, 1999 using the Compton edges of the sources listed in Table 4.6 (left) and the residuals of the fit (right). The additional point at 1.94 MeV is the 3BK ground state, but this was fit with a large 38mK ,û 'background' and subject to systematic errors; this point was therefore not included in the fit. The stdi$l lines in the residuai piot represent the uncertainty in the calibration.
(243.2k0.9 &/MeV) x T, Channel = (38.8*0.6 chan) +
1 + T-df(2.50~0.55)xl~2 MeV]
FIGURE 4.24: Non-linear fit to the scintiilator's calibration using the Compton edges of 7 sources. These calibration points and those of Figure 4.23 were taken at mirent t h .
Figure 4.25(b); see below). Additionally, the result quoted by Cliflord, et. al 1431 is
included for cornparison; as indicated by the design simulations, it was hoped that
TWNAT'S scintillator would have a resolution comparable to Clifford's, and it c m be seen that this has been accomplished. At least over the range of the 7 sources, the
resolution follows the square-root law expected if photon statistics dominate uncer-
tainties in the scintillator's energy measurement. The fits actuaily tended towards a
negative offset, i.e. usin - d x - O,; as this is unphysical, the fit was forced
to og = O if it converged to negative values. The width of the Compton edge at
lower energies is more dependent on multiple scattering effects as the y cross section
is increasing; this complicates deriving the resolution for 13?Cs and the annihilation
radiation from 22Na, and tends to overestimate the ad, of these pointa When u, = 0,
the fit value of p will increase in compensation, so that the resolution given in the
figure is not underestimated, though it may in fact be ouerestimated,
Figure 4.25(b) is a fit to the same 7 sources (except 13'Cs \vas not included) when
the spectra were taken off-line in an environment where scattering off nearby volumes
was minimized. The fits to these spectra provided cleaner spectra that could be fit
further below the Compton edge than those taken when the scintillator was mounted
in the detection chamber. The dashed line of this figure represents the fit if the
*'"Arn/Be edge was not included; the solid one, whose results are the ones quoted, is
if this higher energy point is included. The result, which is in agreement with that of
the April 1999 fit, is only slightly affected by the 241Am/Be point due to the relatively
large uncertainty in its value. The resolution of the scintillator is therefore taken to be
Gin = d(1.80 keV) x Tsin- It is this resolution function that is used in subsequent
MC simulations.
4.3.4 Stabilization
As mentioned earlier, the gain of the PMT is stabilized to correct for long-term drifts,
changes in count rates, and temperature variations. The system is schematically
depicted in the lower left part of the electronics diagram (page 123), and further
details about the 'stabilization unit' can be found in reference [621.
-4 blue LED is rigidly coupled to the light guide of the scintillator just outside the
(a) April 1999 (b) November 1998
FIGURE 4.25: Resolution of the scintilator as determinecl by fits to Compton edges. The 38K and 2 " . h / ~ e points are not included in (a) because their Compton edges are on top of a large background (ps and neutrons respectively) . The result quoted in (b) includes the 2 4 1 h / 8 e point (solid line) and does aot change the result much from just using the lower-energy ys (dashed line in (b) and result given in (a)).
vacuum system as well as to a photodiode (PD) via a fibre-optic cable. The intensity
of the LED is locked by the stabilization unit which adjusts the power driving the LED
such that the signal in the PD is maintained constant. The temperature dependence
of the PD'S gain was measured to be -O.l%/"C, so large temperature changes would
not be properly corrected for. Although TRINAT uses a temperature stabilized clean
room (f0.5"C to help keep temperature drifts in the Ti:sapphire laser intensity from
extending beyond the feedback loop's range), we decided to house the PD in a casing
stabilized with Peltier coolers to f 50 mK. Thus the LED light output is constant at
the 1 x l ~ - ~ level.
The gain of the scintillator is monitored by the unit and stabilized by locking the
LED pulse height in the PM% this is done by adjusting the (nominally -1850 V) high
voltage applied to it. The 450 nm wavelength LED was specially chosen (versus more
FIGURE 4.26: Stabilization test of the scintillator's gain. Plotted I the peak posi- tion of a * 0 7 ~ i source in the scintillator over the course of hours. The system can be seen to keep the gain of the PMT constant even when subject to large count rate changes and temperature variations.
0.950 -;
readily available green LEDs) to closely match the spectrurn of the BC408 scintillation
light as well as to overlap the peak of the photocathode response (both 425 nm). This
is important to ensure that the spectral sensitivity of the photocathode to the LED and scintillation light be as closely matched as possible. Also, since the temperature
dependence of the photocathode's quantum efficiency is a function of the incident
photon's frequency, good overlap ensures that temperature variations are properly
corrected.
Tests of the stability were made using the electron conversion peak in 207Bi. The
scintillator, unlike the silicon detector used to test for energy straggling, does not
resolve the two iines; the spectrum has just one peak at roughly 1 MeV. The 207Bi
decay also has a few y lines, but a coincidence condition with the DSSSD removes
this background; the fuil absorption peak of the mono-energetic positrons is a nearly
Gaussian distribution, which is a good monitor of the PMT gain. The results of one
of the stabilization tests, when the detector was subjected to large changes in the
count rate as well as the ambient temperature, is given in Figure 4.26. The rates were
rtandard deviation of points = 0.40%
1 5 r
O 1 2 3 4 5 Time [hrs]
O 1 2 3 4 5 Time [days]
FIGURE 4.27: Long term test of the scintiilator stabilization system. The count rate was a constant 750 Hz and temperature variations are las t h a . 0.5'.
changed by introducing y sources ('37Cs and aCo) which do not contaminate the
electron conversion peak again because of the DSSSD coincidence. In the experiment
the count rates only Vary from 100 - 500 Hz, predominantly due to changes in the
amount of 38K background. The test dernonstrates that, after an initial jump, the
stabilization system adequately corrects the PMT's gain when subject to count rate
changes 5x larger than we see in the experiment.
The temperature dependence of the system was investigated by adjusting the
ambient temperature of the TRINAT laboratory by 3"' which again is much larger
than experienced during the correlation experiment. As it did when the count rate
was tested, the stabiization system recovers shortly after an initial jump, and correctly
locks the gain.
The overaii stability of the gain including large count rate and temperature changes
is 0.4%. With only small, long-term drifts in the temperature and a constant count
rate of 700 Hz, the long term stability was tested over a period of four days Mth the
results given in Figure 4.27. Though the gain shows a dehi te drift of approximately
-O.4%, the short term variations in this case are seen to be as low as 0.05%. The
Compton calibration has an uncertainty of 0.14% (extrapolated to 5 MeV) so the
stabilization unit corrects the gain well enough considering how weii we can measure
the energy.
Approximately every three days during the April/May 1999 nui, calibrations using
nNa and 60Co were made so as to monitor the long term drift of the scintillator
gain. The differences in these calibrations are only about as large as the uncertainties
in the fits themselves, and so the system is stable to within our ability to measure it
this way. Increased sensitivity to long-term drifts rnight be obtained by comparing
the scintillator's on-line ,û spectrum as a function of time. Indeed, as we will see
in the next section, the final calibration of the scintillator is taken from fitting the
/3 spectrum of the entire data set. By doing this, we become less sensitive to the
long-term drifts because in this case, the calibration is fit to the gain averaged over
the entire running period.
4.3.5 38mK Results and an Extended Calibration
The on-line spectrum of the scintillator on its own, despite the 75 cm collection-
detection trap separation and the lead shielding (both of which help greatly), is still
dominated by the 38K background. The y-@ ratio in the pre-scaled scintillator spec- trum was measured to be 45:l by simultaneously fitting the Compton edge of the
2.2 MeV 7 and the 38mK /3 spectrum to GEANT simulations of the two. The hard-
ware DSSSD coincidence greatly reduces the ground state background by a factor
of 35, but the y events cannot be totally removed this way because t hey do in fact
have a non-negligible probability of firing both detectors in the telescope, The most
likely mechanism for these -y coincidences is when the photon Compton scat ters in the
DSSSD and the collisional partner, the electron, is cletected in the nearby scintillator;
both detectors in this case register an event which is thus inàiscernible from a (good)
B event.
Here and throughout the rest of this section, the scintillator spectra presented
require a coincident DSSSD event that passed the analysis scheme outlinecl in i4.2.5. This condition, which can also be imposed in simulations, is used because it provides a
cleaner ,û spectrum by reducing the 38K background; however, the 2.2 MeV ys do fire
both detectors and so will d l need to be included in the analysis as a background.
The shape of the on-line /3 spectnun, particularly the iow-energy part of it, was
founci to have a strong dependence on the conditions imposed by the DSSSD anal-
ysis scheme. In order to get 95% of the events in the resolution functions of the
strips, a a,,, = 3 energy acceptance should be chosen (see Equation 4.5 on page 57).
By definition, the variation of the energy reading in the A E increases with the
AEMssD 5 SaDsssD acceptance, and the suppression of background events will be less efficient; note also, however, that it tends to accept more higher-energy DSSSD
events (2 500 keV) since these events are the ones which suffer most from inconsis-
tencies between the strip calibrations (see for exarnple Figure 4.13). More precisely,
a tight cut of ~ u D ~ ~ ~ ~ will aclude (goodt) higher-energy DSSSD events whereas the
larger acceptance retaàns them. The plots in the top of Figure 4.28 show the Kurie
plots$ for the different DSSSD energy agreement conditions of luosssD, 2oDsssD and
30DsssD. The tightest condition, AEDsssD 5 la^^^^^, ~ C C O U ~ ~ S for 68% of the 3aDsss~
overali acceptance; the (1 - 2)UDsssD and (2 - 3)uDsssD events contain the additional
26% and 6% respectively. In order to compare the spectrum shapes (bottom of the
figure), the spectra have been renormalized so t hat they al1 have the same totd num-
ber of counts and then the difference from the average was taken. The spectrum
shape of the 2~~~~~~ events is not dramatically different From the tight ~ u D ~ ~ ~ D cut,
but does have slightly more events at lower 0 energy. The 3sDsssD events have a
very different shape which is e-upected, whether the loi-energy events correspond to
hi&-energy DSSSD events or to a background; 84.4.1 discusses the total (E + 4E) energy reading which discriminates between these two cases. -4s statistics is not a
concern in the P spectra from the April/May 1999 nin, a DSSSD energy agreement
condition of lapsssD is used throughout this thesis unless otherwise noted.
GEANT simulations of the 38mK decay were programmed [63j to generate the initial
positrons (as well as the recoils) from the trap* according to the decay rate, Equa-
tion (2.24); the Standard Mode1 values of the Cs = Cb = O and CV = C l = 1 are
(always) assumed. The ps are emitted isotropically into 47~ and tracked until they
either annihilate or exit the detection chamber. h y generated secondaries are also
Eully tracked to ensure the simulation properly accounts for coincidences where the
+'Goodt in that it was an e* which fird the DSSSD; it will be 'baà' in that we wiU have a large uncertainty in the overaii energy,
*The response function is not accounteà for in Figure 4.28 so as to show its dects. 'The finite size of the trap, as measured by CCD cameras, is incorporated in [631.
FIGURE 4.28: Scintiilator Kurie plots for different DSSSD x- and y-strip energy agreement conditions. The Kurie plot with AEDsssD 5 3Cq)sss~ follows a straight line above the 2.2 MeV 7 background; the events above the end point (nominaily 4.75 MeV, which is below the Q-value because of the energy lost in the Be foil, DSSSD and Teflon) are from Compton sumrning of the annihilation radiation. A DSSSD energy agreement condition set between (2 - 3)cQsssD does not contribute many events compared to the rest (top), but have a much larger low-energy tail (bottom; renorm&zed).
scintillator and/or DSSSD only fires as a result of the annihilation radiation or a 6 ray
that goes kom one into the other. The scintillator and DSSSD singles spectra are in-
cremented by the total energy deposited in t hem by eit her the primary pwit ron or any
generated secondary. If the simulated energies deposited in the detectors both pass
their respective t hresholds, as defined by t hose used in the experimentd data analysis,
t hen the telescope's singles (i.e. the Tp = Tsin + EDsssD coincidence) spectrum is also incremented.
Simulations of the 38K decay were made in an attempt to generate the shape of the
y background. The MC generates the 2.2 MeV photons at the collection trap since
this is where most of the ground state's activity is concentrateci. In order to speed
up the simulations, the 7s are not emitted isotropically, but are initially directed
towards the 0-telescope in a cone that covers the area exposed by the 15.24 cm
diameter detection chamber. The Monte Carlo does not contain any volumes outside
the detection chamber, but it is important to start the .ys from the first trap because
they can (predominantly forward) scatter in the chamber wall and this probability
will depend on the initial direction of the y. The lead shielding is not included because
the CPU requirements would be exorbitant; this will certainly affect the simulated
38K spectrum shape because the ys that (fonvard) Compton scatter in the lead (and
then detected in the telescope) wvill appear as a source of ys that are distributed with
energies less than the original 2.2 MeV. The Kurie plots of the scintillator's on-line @ spectrum in Figure 4.28 c m be seen
to be linear above about 2.2 MeV. The deviations are ptedominantly due to the 38K background, but effects from the response function of the scintillator must also be
included. The response function effects are most clearly visible as the high energy
tail near and above the end point, and by the large number of events just above
threshold; the former is from Compton summing of the annihilation radiation while
the latter arises from events where the positron was stopped in the DSSSD and one
of the annihilation quanta Compton scattered in the plastic. Bath of these effects are reproduced in GEANT as can be seen in Figure 4-29; this is a fit of the data to MC simulations of the 38mK spectrum and the ground state's (coincident) 7 background.
The Compton edge at 340 keV corresponds to the 511 keV annihilation quanta and
has in fact allowed us to keep the offset and the slope of the energy calibration
A = 289.5910.0 &/MeV x2/v = 1.tû64 CL. = O
FIGURE 4.29: Fit of the amK spectrum to a Monte Car10 simulation that includes the 2.2 MeV background. This spectnun represents the energy deposited in the scintiiiator (only) detected in coincidence with an event in the DSSSD. The clasiictl line represents a simulation of the /3 spectrum and the 6iied curve is the 38K gound state background. The sdid line is the sum of these two simulations over the fitting range (304 - 5204 keV, over the entire spectrum).
both free to vary in the fit because we bave two distinct points in the data: this
Compton edge and the known Q-value of 5.022 MeV. Note that the /? spectrum only
appears to extend al1 the way out to the Q-value because of Compton summing of
the annihilation radiation; the bulk of the missing energy is deposited in the DSSSD and will be retrieved in the telescope's total energy reading. Approximately 25 and
40 keV are lost in the P window and the Teflon wrapping respectively; however, these
smaii dead layers are included in GEANT and so this energy loss should be accounted
for in the simulations.
In addition to the slope and offset of the energy calibration, the relative normal-
ization of the 38K and 38mK Monte Carlos and the resohtion of the y spectrum were
aiso free to vary in the fitt. The resolution of the p spectrum was h e d as the function
obtained from the Compton edge analysis (Figure 4.25). The same could not be doue
for the 7 spectrum because the Compton edge ends up being much too sharp. As
mentioned earlier, the lead shielding (and other additional scattering volumes) will cause the 38K background not to be the result of only 2.2 MeV photons, but a dis-
tribution of 7 energies; this will smooth out the Compton edge at Exin = 1.94 MeV.
The resolution of the background was therefore free to grow as large as necessary to
get essentially the same effect and indeed it tended towards rnuch larger resolutions
(typically a:E J JG). This method of smoothing the simulated "K spectrurn
to account for the lack of scattering volumes included in GEANT'S geometry allows the
fit to be extended al1 the way down to threshold, approximately 300 keV as indicated
in the figure (also, see below). The smoothed background is not a perfect approxima-
tion to the true ground state background as is irnmediately evident by the fact that
it extends above 2.2 MeV, the total energy of the initial photon. The calculated 14% contamination from the ground state represents an additional 9 . 4 ~ suppression from the DSSSD analysis scheme (using a AEDsssD 5 laDsssD acceptance) on top of the
35x suppression from the hardware scintillator-DSSSD coincidence. This calculation of the amount of 38K background present in the /3 spectrum can only be considered
approximate, however, because the fitting of the relative normalization is based on the total integral number of counts in the two MCs. This will depend on the shape of the
spectrum and, since it changes rapidly below about 250 keV, is subject to systernatics
which far outweigh the statistical uncertainty quoted.
The imperfect understanding of the -y background is the source of mostt of the dis-
crepancy between the data and the fit, most notably in the 'Valley' at about 450 keV. This compromises the effectiveness of the 511 Compton edge in reducing correlations
between the fit parameters (in particular x,, and X of the scintillator calibration).
The fit of the edge's position in channel number is still good, however, because the
7 spectrum is smooth and does not change very much over the limited range of the
511 Compton edge itself. It wouid be very useful to obtain an experimental shape for
this 38K background because it wouid allow more precise calibrations d o m to lower
energies and provide an overaii better understanding below 2.2 MeV. Attempts were
t.4nother possibiity is an underestimation of the low-energy tail by GEANT, but this d l be shown to be s m d by the results of 84.5.2.
TABLE 4.7: Map of the X2/u as a function of the low-energy cut-off of the fitting region, The threshold determined nom the y sources of 131 chaanels is consistent with this map. The high-energy limit of the fit was fixed at channel 1550.
made to get the shape from events when the 38mK was not trapped, but with only
a 7.6 min haif-life of the ground state, they did not provide nearly enough statistics
in the scintillator-DSSSD coincidence spectnim. A dedicated period of the ne* nin
should be directed towarcis measuring this background, simply by not trapping the
38mK but still having ISAC provide us with the potassium beam.
In fitting the on-line P spectrum, the x2 was found to depend on the fitting region
used, particularly with the low-energy cutoff. This is not surprisingly because the
low-energy part of the spectrum is where the greatest contribution to the x2 cornes
from. In order to have an unbiased estimate of the true energy threshold of the
scintillatort, the threshold was determined from spectra of y sources (88Y, 6 0 C ~ and
13'Cs) which do not contain a 511 Compton edge. These spectra showed that the
threshold does not affect the scintillator's spectrum above channel 131. -4s a check,
the x2 per degree of freedom (v) was mapped as a function of the Iow-energy cut-off
with the results given in Table 4.7. The x2/v as well as the fit parameters can be seen
to change dramaticaliy beiow channel 131, but are ail more or less constant above it
and so this map is consistent with the threshold value found in the y sources. The
threshold is therefore dways taken to be at channel 131 in the scintillator spectnim;
for the calibration of Figure 4.29, this corresponds to 306 keV.
tIt is diff idt to differentiate between the fast fail immediately below the 511 Compton edge and that caused by the threshold.
The dependence on the high-energy cut-off is less sensitive; fits where the upper
limit of the fitting region was varied between 4.94 and 5.46 MeV showed that the
X2/u remained essentially constant: it ody changed by +1.8% at 4.94 MeV, but by
+11.1% at 5.46 MeV. Note that a small background subtraction, not included in the present analysis, would reduce the dependence above the end point; the MC has
virtually no counts above 5.3 MeV, so the X2/u naturally rises due this high-energy
tail (which probably is a result of the cosmic ray background). The fit values for xo
and X (i.e. the calibration) changed only by a negligible amount, f0:$7% and -o,ol 0
respectively. These differences are within the fit parameter uncertainties, so the fit
of the p spectrurn is independent of the hi&-energy cut-off. The counts in the data
are predominantly background above channel 1550 (= 5.2 MeV) and so this value is
chosen for the B fits which follow.
The calibration fit in Figure 4.29 is rather different from that obtained using
the 7 sources (see Figure 4.23 on page 73). The channel numbers corresponding to
5 MeV differ by 2.2%, which is well outside the difference allowved from the calibration
uncertainties. This effect mas bc a result of inefficient light collection from where the
ps generate scintillation light cornpared to the whole; since the ys Compton scatter
hornogeneously t hroughout the plastic, they average over the bulk of the plastic while the ps are concentrated more near the front face of it. Another possibility is the
inaccuracy of GEANT'S ability to properly simulate (à) annihilation-in-flight, (ii) the
differences in the radiative energy losses of e' to that of a Compton scattered e-,
and/or (àia) the relative cross-sections for radiative and collisional energy losses of
positrons. Differences in calibrations from 7 and positron sources has been noted by
others [641, but a full e-xplanation has stili not been found. Since the @ spectrum
fit is to the data themselves and it averages over variations throughout the running
period, this calibration is preferentially used instead of the Compton calibration. The
resolution, however, remains defined by the fits to the Compton edges.
For the p - u correlation experiment, the additional coincidence condition with
the recoil detector virtuaiiy eliminates the 38K Y background and indeed, may one
day be used as a test of the lower p energy part of GEANT'S simulations. At present
though, other uncertainties preclude doing this (see §4.5.2), and so the measurement
on a will be sensitive to how different the 0-teiescope's response function is fiom
GEANT simulations. The present analysis scheme does not, however, use events below
TB = 2.5 MeV; thus the fit of the scintillator's 0 spectrum that is relevant to the
experiment does not include the ground state background. Without fitting the 511
Compton edge, the offset of the calibration cannot be fit, so we fix it to the result
obtained fiom the B fit of Figure 4.29, namely z, c, 42.3 channels. With now only
the slope, A, of the calibration Eree to vary, Figure 4.30 shows the result of a fit
where the low-energy limit of the fitting range was 2.2 MeV. Over this more limited
range, the GEANT simulation of the /3 spectrum is excellent, with a confidence level of
over 80%. This means that for the correlation experiment, GEANT does a very good
job of reproducing the ,!3 spectrum over the region of interest. A potential source
of concern would be a difference in the value of A fit here with that of Figure 4.29,
however the difference of 0.05 ch/hIeV is well within both of their uncertainties.
The two calibrations agree, and this result is important because it shows that the y
background and, to a lesser extent, the DSSSD-GEANT discrepancy, does not bias the scintillator's calibration obtained in fitting the 0 spectrum. The GEANT simulation
of the ground state smoothed with a resolution much worse than measured, is in this
case an adequate approximatian because fitting above it does not change the best-fit
calibration. Therefore, the calibration of Figure 4.29:
c m be considered reliable and should be used in any analysis of the April/May 1999
data set that requires the 0s energy. The differences from the observed DSSSD energy
spectrum and that of GEANT'S may also affect this fit because this energy is subtracted
in both the MC and the data; the calibration may therefore be biased, but only slightly
since the discrepancy is very small compared to the energy deposited in the scintillator.
This effect is addressed later in $4.4.1.
4.4 The Ptelescope
Whereas earlier sections dealt with the E and AE detectors separately, this section
deals with the results hom the April/May 1999 run of the 0-telescope as a whole.
First, the telescope (meaning the T,, + EnsssD = Tp) P spectra are presented and
FIGURE 4.30: Fit of the 3 8 m ~ spectrum above 2.2 bkV to a Monte Carlo simula- tion. The fitting region (2193 - 5205 keV) is above the 38K background contam- ination, and so the MC in this case is only of the P spec tm. The offset of the calibration is fixeci at xo c, 42.3 from Figure 4.29 because the 511 Compton edge is not fit.
discussed. FoHlowing this is i4.4.2 which compares backscatter calculations of GEANT to those on-line using the DSSSD position information. The uniformity of the tele-
scope response to the incident positron's direction is discussed in i4.4.3. The final
section explains how the P-telescope may be used to measure be, the Fierz interference
parameter.
4.4.1 Total P Energy
The previous two sections have provided us with both the calibration and the resolu-
tion of the scintillator and al1 of the brty-eight DSSSD strips, as well as an effective
position decoding analysis scheme. The calibrations allow us to sum up the scintillator
and DSSSD energy readings for a total Tp measurementf. Having an understanding of these detector resolutions further allows us to program GEANT to simulate the
telescopels B spectrum of the 38mK decay; the resolutions are essential because they
must be applied before adding the two energies (as is the case with the real data)
rather than Btting them after, Prior to fitting or comparing the on-line spectra to these MCs, considerable effort must be made to ensure that al1 conditions placed on
the on-line data are reflected in the simulation. As a quick list, the most important parameters to consider are:
The detectors' resolutions and calibrations (fixeci).
1 The scintilIator's low-energy threshold (user parameter).
a The DSSSD low- and high-energy thresholds (user parameters).
The active area of the DSSSD (user's choice; for example, whether or not the
edge strips are included).
We saw earlier in Figure 4.28 how the value of ocut used in the analysis of the
on-line data has an impact on the resul ting P- telescope's spec tra. The correspondhg
figure for the total energy, TB = Tsin + EDsssDl is given in Figure 4.31. -4 qualitative
cornparison of these two figures tells us that the additional low-energy events in the
fact, this coupled with the position information gives us a rneasurement of the positron's momentum, p,.
- al1 accepted ps
-
O 1 2 3 4 5 6
T, [MeVI
FIGURE 4.31: The telescope's ,û spectra for different AEDsss~ conditions. These are the corresponding spectra for Figure 4.28; they have the sarne conditions applied and ody ciiffer in that the DSSSD energy has been added to the scintillator's.
larger AEDsssD spectra are not due to the greater acceptance of high-energy DSSSD events (as suggested they may be in 843.5) because adding back the DSSSD energy
does not remove this low-energy tail. With just a Little imagination, one can see the
Compton edge of the 2.2 MeV background in the 3aDsssD 0 spectrum, indicating
that the larger acceptance instead allows a greater contamination from this, and
presumably other, backgrounds. Thus unless statistics is a great concernt , the tighter
AEDsssD energy cut of < ~ u D ~ ~ ~ ~ should definitely be used.
The Compton edge at 340 keV, prominent in the scintillator's energy spectrum,
no longer appears once the DSSSD energy has been added back in. If the peak
remainecl, it would indicate that these events fired just above the DSSSD's threshold,
and they would likely be due to a random fluctuation above threshold, randomly
coincident low-energy photons, or they might point to a flaw in the analysis scheme.
The fact that the edge does not remain in the telescope's 0 spectrum shows that an
appreciable amount of energy was deposited in the DSSSD; this is consistent with a
tWhich may be the case once the recoil coincidence is applied, for example.
positron scattering or stopping in the strip detector, and then one of the annihilation
quanta Compton scattering in the scintillator. The total energy of these events are
distributeà above the Compton edge in the ,O spectrum, so the telescope wàll contain
some background from these events, but these events are ( i ) impossible to remove
by a re-design of the telescope, (ii) a smali percentage of the total number of events,
and (iz'i) reproduced reasonably well by GEANT for the scintillator's spectrurn as
shown in Figure 4.29. They can be greatly suppressed, however, simply by setting the scintillator's (software) threshold at or above 400 keV in the analyzing program.
Although it is good that these events do not remain a well-defiaed Compton edge in the telescope's spectrum, it also perhaps a little unfortunate because we no longer
have two distinct points to fit to as before. As was the case when fitting above
the 2.2 MeV background, we must fix the offset of the energy calibration. For the
telescope, this 'energy calibration' will be different from the scintillator or DSSSDYs
in that it is alreadg calibrated, and therefore in units of energy rather than channels.
If we were to fit the MC to the telescope's 0 spectrum and if everything was perfectly
calibrated and simulated, the telescope's calibration:
would yield an offset of T; = O keV and q = 1 keV/keV for the slope. The poor
agreement between the DSSSD's energy spectrum and that simulated by GEANT is an
immediate indication that this is not the case. To first order, however, the difference
between the DSSSD and GEANT'S energy distributions is given by the difference of
their mean dues , which is -10.3 keV (see Figure 4.18). We therefore can account
for the AE's discrepancy on auemge by fixing the offset to be f10.3 keV.
The spectrum from the on-line data and the results of a fit to a MC simulation
is given in Figure 4.32. As mentioned earlier, the resolution of the detectors must
be input into the MC before adding the detectors' energies up, and therefore the
resolution cannot be fit. The free parameters in this fit are just the dope of the
calibration and the relative normalization of the P and 2.2 MeV y background. The
simulation of the ground state, which is convolutecl with the same resolution as in
Figure 4.29, is not as good an appravimation anymore as the Compton edge is clearly
visible at 1.9 MeV. The unreasonably large resolution of the ground state in fact
q = l.ûûûiû(t5) keV/keV x2/v = 2.3097 C.L. = O
FIGURE 4.32: Fit of the 38mK Tg spectrum to a Monte Car10 simulation. The offset was fixed at 10.3 keV because of the average ciifference between the GEANT sirnu- lated and measured DSSSD energy spectra. The 38K background is not reproduced as weU when the DSSSD energy is added back in.
can be seen to limit our low-energy fitting range because it unphysicaliy places events
belorv the scintillator's threshold, resulting in a low-energy tail. One point that should
be noted is that the resohtion of the 38K background in the DSSSD was not adjusted
since we bave no idea what the DSSSD's spectrum is in this case. Indeed, simulations
indicate a higher average energy ioss than the 38mK OS, and since there already is a
DSSSD-GEANT discrepancy, the DSSSD's 38K spectrum can play a more important
rote. Perhaps most significant, however, is the fact that in this case the background
shape in both detectors must be Gxed when fitting the P spectrum, rather than open to
approxhating over-estimations of the resolution. It is not too surprising, then, t hat
this total fl energy spectnim is not reproduced as well. Once again, an experimentalIy
measured spectrum shape for the p u n d state is essential for a good understanding
below 2 MeV.
FIGURE 4.33: Fit of the 38mK TP spectrum above 2.3 MeV to a Monte Car10 simulation. Above the 38K background, the MC is seen to reproduce the on-line 3SmK p spectrum very WU.
The fit to the data naturally has a large X2/u due to the ground state background,
but otherwise the fit looks very good. The fit to the re-cdibrated dope is unity within
uncertainties which is encouraging. As with the scintillator's spectrum, we aiso fit
above this ground state to see how that affects the calibration. The resdt is piotted
in Figure 4.33 and again we find that the MC does an excellent job of reproducing the
data. The fits were sensitive to the low-energy cut of the fitting region, and aeeded
to be pIaced siightly hijher than expected (5 2.2 MeV) in order to obtoin good
agreement. The value of 71 is one to within 0.06%, which is a dear indication that the
DSSSD-GEA~VT discrepancy is negligible once the average ciifference is accounted for (TB = 10.3 keV) .
4.4.2 Backscattering Losses
As discussed earlier, one of the reasons TRINAT decided to switch to a plastic scintil-
lator instead of the Si(Li) as the E detector was to reduce backscattering losses. The
low-Z plastic has a relatively low probability for backscattering and the simulations
of Figure 4.2 indicate that Our losses are dominated by backscattering off the DSSSD; this represents an overall loss of event acceptance and can affect the p spectrum if
an annihilation photon then Compton scatters in the scintillator, as discussed in the
previous section.
There is also another type of backscattering loss to consider, and one that can
be measured and then compared to GEANT as a check of how well it is reproduced
in the simulations. Consider events where the ,û first goes through the DSSSD, then
scatters in the scintillator, and eventually scatters back out through the DSSSD again.
These events (hereafter called 'backscattered events') will add to the low-energy tail
of t he telescope's response function because energy is deposited in both the E and AE detectors. The pixel of the DSSSD that the positron enters through will, in general,
differ from the one that it backscatters out through; in this case, the event is flagged
as a double hit and vetoed by the analysis of the DSSSD position information. Vie
assume that these double hits in the DSSSD, where both ( 2 and i ) group hits passed
the position and energy cuts (ucut = 1) of the DSSSD analysis scheme, are primarily a
result of this backscattering effect; the probability of a random double P coincidence
is very smaii considering our P event rate of typically 5 30 Hz. The fraction of these
backscattered events in the 38mK data was 1,12% of the total number of accepted ,û (one DSSSD hit) events using ucut = 1 in the analysis scheme. If the DSSSD energy
agreement is increased to uot = 3, the fraction of backscattered ,û9s increases to
1.35%. In either case, events where there was more than two groups in 2 or @ were
still excluded (les than 0.09%), and not considered to be a backscattered P. GEANT simulations can also have these backscattered events tagged, and the cornparison to
the on-line data gives an idea as to how well GEANT calculates scattering within the
scintillator. The result of a simulation of the unweighted B spectrum was 1.22%. This
is in agreement with the data, considering the spread in the results fiom the choice
of G u t -
4.4.3 Uniformity of Response
The gain of the scintillator may Vary according to where in the plastic the scintillation
light was generated because the light collection efficiency can vary with position. Off- line tests were performed with a *07Bi source where the source was moved along the
length of the scintillator, with variations in the gain found to be at the 1% level.
This preliminary result is encouraging, but the effect on the correlation expriment
can be different. By placing different position conditions on events in the DSSSD, Ive
can use the on-line 38mK data to map out the non-uniformity. The difference in the
B spectra of the different cuts can be used to determine our sensitivity. For example,
if we define an 'inner' DSSSD position cut and and outer DSSSD cut, we cm check
that the light collection of the scintillator around the centre is the same as near its edge. In order to obtain good statistics on the /3 spectra, a large area at the centre
of the DSSSD (defined by a 01.2 cm circle) was chosen for the inner cut; an outer
area defined by strip positions outside the 02.4 cm circle has the same area as the
inner circle. Circles are defined (rather than 0.1 mm \ide squares) because of possible
inter-strip events, even though they are but a small fraction of the the single-strip
events. Note that these circles do in fact have an equal number of the dominant
single-strip pixels (only one strip in 2 and one in e). The /3 spectra of Figure 4.34 are the results where the inner and outer position
conditions were imposed on the 38mK data set. The outer DSSSD events' efficiency is
83% that of the inner ones and this is mainly due to the reduced efficiency of the edge
strips as discussed in i4.2.5. In order to aid in comparison of the spectra, this relative
inefficiency has been corrected for by renormalizing the inner DSSSD's spectrum to
have an equal number of counts as the outer. The plot in the upper right of the figure
is the difference in the number of counts, and generally the agreement c m be seen to
be very good. The difference is < f 5% in the number of counts (For energies less than
4 MeV; see plot in upper right of Figure 4.34) indicating uniformity of the scintillator
and telescope as a whole to the direction of the incident P. Above 4 MeV, the difference 2Q a considerable fraction of the total number oicounts
and we should determine how sensitive our measurement of the correlation parameter,
a, wi i i be to this non-uniformity. To that end, full Monte Car10 (651 simulations with
FIGURE 4.34: Uniformity of response of the plastic scintillator. The two ,O spectra ciiffer by the position condition of the DSSSD imposed on them: the shaded 01.2 cm circle in the ceutre (%mer') and the shaded area of the corners ('outer') chosen such that the two have equal DSSSD areas. The plot insert in the upper right is the (percentage) difference in the nuniber of counts between the two spectra.
both responses must be performed. The resulting fits of a to t hese fake data, and
their deviations from that input into the MC, will give us the relative importance of
the telescope's uniformity.
4.4.4 The Fierz Interference Term
The Fierz interference term, b,, of Equation (2.24) will, in general, be non-zero if
a # +l; this is made explicit by Adelberger, et. al [IO) when they express their mea-
surement in terms of à r i+be(ze,Ee). The final analysis of TRINAT'S ,û - v cor-
relation experiment may very well incorporate 6, by directly fitting the CS,V and
C$,v parameters; in the present analysis scheme, our sensitivity to this interfer-
ence term will be small because it only considers events above Ep = 2.5 MeV where
0.093 < mJE, < 0.17. With the limit of Ibe( 5 0.007 1111, the impact on the correla-
tion parameter is relatively mail: ü would only differ from a at the 0.1% level.
The abity to reproduce so well the unweighted B spectrum for the 38mK data
aliows us to consider fitting be using the telescope on its own. A preliminary investi-
gation into this prospect has been completed which looks very promising, but a more
detailed analysis is necessary to compete with current limits. The unweighted P spec- trum was fit to MC simulations as in i4.3.5, but this time the number of counts in
this MC spectrum was renormalized
and Wyld's Coulomb correction):
Ni,c(E)
by (see Equation (2.24) with Jackson, Trieman
where be was a free parameter in the fit. The other parameters left free to vary
were the slope of the energy caiibration (A of Equation (4.10)) as well as the relative
normalization and resolution of the 38K background (which \vas not renorrnalized by
Equation (4.19)). The slope of the calibration in t his fit agreed with the fit when b, was set to zero: X = 289.59 f 0.13 versus 289.72 f 0.15 channels/MeV and xo = 4Z2 f 0.4
versus 42.3 f 0.4 channels). This is an important result because it shows that the Fierz interference term will not bias our on-line p calibrations of the scintillator.
The normalization of the y background did significantly change, however, and was
highly correlated with 6, (97%). This is a result of the poor spectmm shape of
the background simulated by GEANT, and limits our sensitivity to be since energies
below the Compton edge are where the interference term will have the largest effect.
The result of this fit was 6, = -0.05 & 0.03 which is much larger than the limit of
Ibe[ 5 0.007 f 0.005 [Ill reported earlier. If only A of the calibration and b, are left free (with the background fixed to the results of the fit with 6, = O), then the limit
on b, is very good and comparable to the published results: be = -0.003 k 0.008. There is no justifiable reason for Luing the background parameters, however, and this rcsult is only an indication of what limits may be attainable once efforts in
understanding the shape of the 38K background are cornplete; in addition, questions
about non-linearities in the scintillator gain will need to be addressed, and calculable
higher-order corrections must also be included.
4.5 P A r Coincidences
The MCP-@-telescope coincidence spectra are presented in this section. As this is
largely the basis of A. Gorelov's thesis [la], the analysis contained here is preliminary
and only meant to give an indication as to how the /3-telescope will be used (as well
as how it will perform) in the ,8 - v correlation experiment. The k t section explains
the scattering effects that are present in our geometry and shows how well GEANT reproduces them. The other section compares simulations of 0 spectra where a recoil
coincidence is required to those observed on-line in April/Way 1999.
4.5.1 Scattering Effects
The plot in Figure 4.35 is a Monte Carlo simulation of the 0-v correlation experiment
for the Ar" recoils. This MC is a combination of GEANT (positron tracking) and
A. Gorelov's (recoil tracking) code [651. The electric field of -829 V/cm is not strong
enough to collect all of these recoils, but it does increase the eficiency greatly and is
easily implemented in the analysis if it can be considered uniform, as it was in the
simulation. The efficiency of the MCP is assumed to be uniform, and its active area
has been assumed to be 02.5 cm. The kinematic limits indicated in the figure were
calculateci for the back-to-back geometry using point-like detectors wit h an ideal 0 response function. The Monte Carlo includes the finite trap and detector sizes, as
well as the energy and timing resolution of the @-telescope; the Compton summing of the annihilation radiation is clearly evident as a cidge at P energies above the slow
branch.
The events outside the kinematically allowed region at longer times-of-0ight de-
serve special attention because they correspond to positrons which were not emitted
toward the D-telescope, but (predominantly back-) scattered before entering; this will
bias the initial direction of the recoil and hence the TOF. Consider the case where
the At' ion recoils toward the B detector; the electric field is strong enough t hat it
will be tunieci back and accelerated onto the MCP. If the decay is in the fast branch
of this 'reverseci geometry', then the recoil TOF dl be constant (at N 1 ps) and
the leptons will both be emitted toward the recoil detector. The positron then has a non-negligible probability of backscattering off the MCP (lead glas) or one of its
electrostatic hoops (aluminum) into the ,8-telescope. If this happens, the event would
be detected just Like a 'good' ,8 - Ar coincidence, but outside the kinematic limits at
longer TOF. These events are plotted separately in the bottom of Figure 4.35- Now
FIGURE 4.35: Monte Carlo simulation of the recoil TOF as function of P energy for ~ r + ' recoils (top) with the kinematic ümits for the back-to-back geometry (solid lines), Events in the kinematicaily forbidden region at longer times-of-Elight are a result of positrons that scattered before firing the Ptelescope. Most of these events are a result of the p backscattering off the recoil detector or one of the electrostatic hoops (bottom).
n
a 900 - Y
Er. eoo - '
H .I
O 700
Q) L
=-., . . x. . . . . . . '. : . \., . . . . .
1. :.: \ '
. . . . . . ? .,;. . . . . . . . . . . . ..:. . . . . . . . . ....... ::..\ ...... .......... : :Y,.
- : : : : : : ; : : : : . ; : ::::::::>-.' ............ ..\.. ....................... . v . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ............................. -5' ................................. s,. . ..................................... -
-
500 1 , , 1 r O 1 2 3 4 5 6
T, (MeV1
-
*
1100 -'
1000 -. - 2 9oo Y
kt
800 - e .d
& 700 - L
600
I , t I
backscattered off MCP or hoop
.,:: .i:ii*#yjijli;;ii;ii:i:i - .*- .:.: i: : .................... .t..:'....*....... . . . . . . . . ........ . . ............. :'::' - : .............. . . - ' . - . ........ . . . . . . . . . . . . . . . . . ...... ....... . . . - . . ..y..- ........ . - . - . <..: . . . . . . . . . . . . . . . . . . . 7 * * . . - : . y - . . . . * . . . ..
',
--.... . -- . . . . A ., ... . . -. .
-& ... :. .... - .
consider a backscatteredt /3 which came from the slow branch; the recoil in tbis case
is directed toward the recoil detector rather than away, and so the event is detected
at faster TOF. Essentialiy, the s p e c t m shape of the backscattered ,f3s will have the
same form as the 'direct' ones, but it is reversed in TOF and shifted down in /3 energy
(from the energy lost in the volume where it backscattered).
The events where positrons backscattered off the MCP or one of the hoops were
tagged in the Monte Car10 once it was found to be a clearly visible background.
Table 4.8 gives the percentage of /3-Ar coincidences where the positron scattered off
a volume before firing the telescope. The G10 frame of the DSSSD and the inside
part of the front face of the telescope vacuum chamber that defines the /3 window are
seen to scatter the largest fraction of the events. Exclusion of the edge strips does not
reduce these scattered events as much as one might expect; t his agrees with the lack of
(real) low-energy tails observed in the edge strips, although other systematic effects
still favour restricting the DSSSD active area. The major concern regarding these
events is that though they will appear to be good /3 events, they will be registered in
the wrong DSSSD position (and so will yield an incorrect measurement of Bpv) and will be in the low-energy tail due to the energy lost when scattering. The MCP and
electrostatic hoops are the dominant source of backscattered 0 events and explain
most of the background seen above the slow branch in TOF; additional volumes off
which the B can backscatter include the rnount for the hoops behind the MCP, the
MW's flange and the vacuum chamber itself.
We do in fact see evidence of these backscattered Bs in the on-line data, rvhich is given in Figure 4.36. -4 comparison of this with the simulation of Figure 4.35 shows
that the two are very similar and that the MC does a good job of reproducing the scat-
tered events. The two differences between them are a srna11 constant background from
random MCP-telescope coincidences, and the tail of AP2 events at TOF 5 550 ns.
The TOF for the fast branch of the backscattered events of the Ar+* will overlap the
(direct) fast branch of the Ar*'. This same effect will occur for a11 higher, otherwise
resolved, charge states as weU. Once a EB cut of 2.5 MeV is applied, however, the
backscattered peak is greatly reduced, and the different charge states are separated
+This definition of 'backscatter' is not the same as in 84-42; in this case, it cornesponds to positrons that were initiaüy ernitted in the -2 direction.
p scatterhg volume / @ Ar" h+2 h i 3 h + 4
Steel around 0 window
- including edge strips - excluding edge strips
- excluding edge strips / 1.32% 1.65% 1.85% 2.02% 2.02%
1.85% 2.59% 2.69% 2.73% 2.72% 1.73% 2.27% 2.47% 2.48% 2.42%
G10 DSSSD mounting - including edge strips 1.42% 2.03% 2.28% 2.45% 2.55%
Lead glassofMCP (including edge strips)
TABLE 4.8: GEANT calculations of the percentage of 0s that scatter before firing the telescope for various Ar recoil charge States.
0.28% 0.12% 0.17% 0.19% 0.20%
hoop (including edge strips)
in TOF so that this is not a serious concern.
-4 measure of the quality of the GEANT simulations can be estimated by comparing
the TOF projections of Figures 4.35 and 4.36. Such a comparison for the ArfL recoils
is given in Figure 4.37 where the background was fit to the events between the Ar+' and Ar0 peaks, and the fit cuve (Monte Car10 + background) is normalized to have
the same number of counts as the on-line data. The MCP positioning is not fully
understood at the present time, and known non-uniformities in the electric field of
the April/May 1999 data are not included in this preliminary analysis. However, once
they are, the comparison over the peak of the fast branch should improve greatly. The
backscatter peak is nut quite as large as in the data, but the MC does account for
most of it; the addition of the ceramic rods used to mount the electrostatic rings into
GEANT'S geometry should reduce this difference even furthet ir. future simulations
that include them.
Regardless of the outcome as to GEANT'S reliabityin its absolute scattering rates
once a detaiIed anaiysis is completed, it has already helped improve the experiment
0.48% 0.78% 1.00% 1.23% 1.44%
FIGURE 4.36: Scatter plot of recoil TOF versus Tg Crom the on-iine 38mK mea- surement. The sarne kinernatic limits as in Figure 4.35 are plotted for comparison.
FIGURE 4.37: TOF projections for the and comparison to GEANT. The solid line represents the full GEANT simuiation and the i Iast I I . ( i line are events where the positron backscattered off of the hiCP or one of its electrostatic components before entering the Ptelescope.
L
- i '
-
4 1 I 1
10' - CEANT simulation - ..
***.
- - scattered off MCP or hoop .-. -- - - constant background . *-.. . 1
102 ;
IO' ! -
loO
10' I 'Z I
. * , * '
- a 1 '1
4 l
1 i
I i 1.- ._ ------- ..-------z-=-- -------- t t
400 600 800 1000 1200 1400 Time-of -f light [ns]
by identifying the backscatter background and, once known, its TOF bas proven to be useful in understanding the electric field.
4.5.2 3 8 m ~ /3 - Ar Coincidence Spectra
Although not easily evident in the scatter plots of the previous section, the @Ar coincidence condition virtually eliminates the 38K background. This can be clearly
seen by the examining the recoil-coincident ,LI spectra and looking for any sort of bump
at around 1.9 MeV. Using the same analysis as was done for the ,ü spectra in i4.3.5
and i4.4.1, these spectra are given in Figure 4.38 for the different recoil charge States
Ara, ArL and A3; the Ar3 and Ar4 are continued in Figure 4.39. Both the scintillator
(left) and the scintillator + DSSSD spectra (right) are comparedt to MCs. Happily, there is no evidence of a 38K background in any of the spectra; this result
is expected because the probability of the y firing al1 three detectors is &remely
small. In fact, the MCs generally agree very well with data for Txin x T', < 2 MeV. It is for this reason the claim was made (see page 83) that the discrepancy in the
scintillator's /3 spectrum was predorninantly due to the 38K background, and not from
a grossly incorrect GEANT simulation of the low-energy tail. Note in particular how
GEANT correctly reproduces the Compton edge from the 511 keV annihilation quanta
in the scintillator spectra; this indicates that the discrepancy in Figure 4.29 is due
predominantly to the misunderstood 38K background, and not GEANT'S simulation of the background.
We can actually hope to one day test the low-energy tail of GEALW'S response
function by seeing how well it reproduces the data of these background-free spectra.
At present, however, the data are subject to a number of uncertainties:
0 The MCP efficiency. This is most notably true for the ,4r0 because their ef-
ficiency is unknown and may vary significantly. The charged recoils are al1
accelerated to energies where the response is flat to within 2%.
0 The electric field. Considerable effort has been made to understand the electric
field as thoroughly as possible [57j, but the preliminary analysis presented here
+Note, they are not, in any way, fitted; the relative normalization is chosen so the data and MCs have an quai number of counts.
FIGURE 4.38: The 38mK P spectra coincident with @, Ar+' and .Ar+Z recoils. Overlayed is a Monte Carlo simulation which assumes a uniforni MCP efüciency and a uniform electric field of -829 V/cm. The caiibration was taken from the fit of the unweighted @ spectm, Figure 4.29 on page 82.
F~GURE 4.39: The 3 8 m ~ P spectra coincident with and Ar'' recoils and corn- parison to Monte Carlo simulations.
bas not incorporateci k n m non-uniformities. indications are that the tield was
dose to -829 V/cm near the trap, but decreased to about -780 V/cm at the
surface of the MCP.
The active area of the MCP. The position spectrum of the MCP is not Fuily
characterized yet and this affects the present àiscussion because, althougti we
are integrating over the whole area, we need to accurately know what that
area is. It was faund that by changing the MCP radius by 1 mm the resulting
spectra were noticeably different, especially for the neutrals and lower (< 3)
charge states.
Considering these varied systematics, the present comparison of GEANT to the
on-line data must remain relatively qualitative. The different charge states were
chosen by simple cuts in TOF; a more detailed analysis should corisider the overlap
of higher charge states. To iudicate the differences that currently exist, tables of the
first through third moments of the two spectra are given in each plot of Figures 4.38
and 4.39. The differences are largest for the Ar0 data which is most likely a result of the MCP eEciency. Aii of the charged recoils have better agreement which is expected
since the uncertainties in the electric field are less than the uncertainty in the efficiency
of the neutrals. As we go to higher charge states, the coincident-P spectra look more
and more like unweighted /3 spectra and the agreement generally improves.
Once the systematics are better understood, a detailed analysis of the recoil- coincident B spectra may be made. This would complement fitting the ,&singles
spectrum down to threshold once the shape of 38K ground state background has been
meaured.
Conclusions
A j3-telescope consisting of a double-sided silicon-strip detector and a plastic scintit
lator has b e n designed and constructed ta observe the positrons emitted in the P decay of 38mK.
The energy calibration of the plastic scintillator has been accomplished using the
Compton dges of various y sources as well as by titting the on-line P spectrum to
detailed Monte Car10 simulations. The resolution of the scintillator is derived from
the Compton edge analysis and found to foUow the square-mot law expected if it
is dominated by scintillation photon statistics. The width, ust of the scintillator's
timing resolution relative to a micro-channel plate detector bas been measured to be
one nanosecond. These characteristics fuifill, and in some cases even surpass, the
specifications we had set wben designing the scintillator.
The calibration and resolution of each strip of the A E detector have also been
determined by using both low-energy photon sources as well as the on-line data
themselves. The strip detector is an essential cornponent of the 0-telescope because,
in addition to providing the position information needed for a measurement of the
positron's momentum, it is an effective tag for ,û events and greatly reduces y back-
grounds.
Monte Car10 simulations using GEANT are generally in good agreement Nith the
measurements of TRINAT'S April/May 1999 38mK data set for observables that are
independent of any scalar interactions which will be determined in the final analysis.
There is a (EDsssD) = 10.3 keV diirepancy between the MC and the measured en-
ergy spectrum in the strip detector that is not presently understood, but the impact
on the total Tg measurement was found to be negligible. The sirnuiations reproduce
CHAPTER 5 CONCLUSIONS 107
the scintiUator7s B spectra very well, although our understanding below 2.2 MeV is
complicated by the large 38K (ground state) y background. The E - AE coincidence
reduces this background by more than 99.5%, but there is still about a 15% contam-
ination in the telescope's 0 spectrum. GEANT colculations of the fraction of /3s that
backscatter out of the plastic and through the strip detector a second time are in
agreement with those found in the data.
The goal of TRINAT'S 38mK B decay experiment is to make a precise measurement
of the 0 - u correlation parameter a. The P-telescope's most significant contribution
to the uncertainty in this measurement, a,, is 0.05% taken from the uncertainty in
the 0 calibration of the scintillator; possible non-linearities in the gain may increase
this uncertainty to 0.5%. The uniformity of the scintillator's response over the active area of the strip detector has been investigated, and small differences are noticed.
This is a concern because it will be highly correlated with Op,, but an estirnate of the contribution to a, still needs to be performed. The ground state's y background,
prevalent in the telescope's unweighted @ spectra, is virtually eliminated once a recoil
is required to have fired the micro-channel plate detector. These Ar-coincident P spectra have been compared to preliminary MC simulations and though the spectra
are reproduced reasonably well, further detailed analysis will improve the agreement;
in particular, understanding the active area of the micro-channel plate and proper
inclusion of the non-uniform electric field are necessary for the a measurement.
The data set from April/May 1999 has enough statistics for a 0.3% measurement
of a, but the final analysis is still in progress. Continued experiments are planned with
an improved geometry for a cleaner measurement wit h bet ter statistics. Alt hough the
energy and timing measurements of the positron in this data set are precise enough
considering the statistics accumulated, further improvement in our understanding of
the P-telescope is of course desirable, and below we outline how this rnay be accom-
plished.
The greatest advancement may be made by experimentally determining the shape
of the 38K background and by reducing the amount of this background relative to
the 38mK B spectrum. According to GEANT simulations, the plastic scintillator is
much larger than is necessary for the experiment; by simply making the scintillator
half the current length, we should be able to reduce this background by a factor
CHAPTER 5 CONCLUSIONS 108
of two. The shape of this background c m easily be determineci in TRINAT'S next
experiment by accepting ISAC'S potassium beam but not trapping the isomer; we can
then measure the spectra of just the ground state where both the E and Al3 detectors
fire. Such a determination would aid in calibrating the scintillator and is essential if
the unweighted @ spectrurn is to be used to measure the Fierz interference parameter,
be The need for the large corrections in the extended calibration of the strip de-
tector may be reduced with a better initial calibration of the strips, particiilarly if
this calibration covers more of the experiment's energy range. Calibrations using an
open electron conversion source would complement the low-energy photon sources and
may help to explain the small discrepancy between GEANT'S simulation of the strip
detector's spectnim with that observed in the data.
Possible non-linearities in the scintillator's energy calibration should be investi-
gated further using y sources over the energy range 2 - 4 MeV. Additionally, ex-
tending the Compton edge calibration to higher energies may reduce the differences
observed in the 7 and P calibrations of the scintillator.
Although GEANT seems to reproduce the experiment very well, a complete un-
derstanding of the telescope's response function can only be made using a well char-
acterized beam of mon~energetic positrons. Such a study could be undertaken at
a dedicated facility such as the pelletron facility at the Mau Planck Institute at
Stuttgart [661, or at TRIUMF as has previously been done [401. The appeai of Stut tgart
is the excellent beam quality whereas doing it locally would allow us more time to
optimize the set-up and reproduce the environment of the correlation experiment as
closely as possible.
Response funct ion of the scint illator
Ideally7 the signal generated by the scintillator, XADC, is a perfect representation of
the energy deposited in it, Tsin, In reaiity, every detector has a response function
which relates the actual energy deposited and the resulting observed signal. Typical
response functions (shown in Figure 4.1) have a large peak corresponding to the incident particle's kinetic energy as well as a low-energy tail where part of the energy
loss wvas undetected. The positron response function has an additional high-energy
tail due to the finite probability of detecting the annihilation radiation as well.
To understand the characteristics of the response, we need to understand the pro-
cess by which the particle is detected. The incident particle interacts with the scin-
tillator and suffers energy loss through Coulomb interactions with atomic electrons.
These electrons are then excited to a higher energy level in the atom (excitation) or,
if the energy transferred is greater than the binding energy, is ejected into the contin-
uum. In the latter case, these freed atomic electrons (which are known as 6 rays) suffer
energy losses and excite other atoms as they transverse the scintillator, just like the
primary particle. The atoms in the excited states rapidly decay to the ground state
by fluorescing, which is to say they emit (visible) photons. Some of these photons
are collected on the photosensitive surface of an optically coupled photo-multiplier
tube (PMT), ejecting (at most) one photoelectron per photon. These photoelectrons
are accelerated and focused by electrodes called dynodes, which are typically made
of materials with a high secondary electron emission probability so that the cascade
is multiplied as well. The resulting pulse after repeated amplification through the
dynode structure of the PMT, is the h a 1 signal generated by the scintillator.
Saturation Effects
The efficiency of the excitation process is not unity and in fact is dependent on the
local energy los. Since there are a 6nite number of atoms near the ionizing particle,
it is ceasonable to expect that as the dE/dx gets large, the efficiency is reducecl as
there are fewer atoms nearby that are still in the ground state. This saturation effect
was first investigated semi-empirically by Birks [671 who parameterized the amount
of scintillation light generated (luminescence) in terms of the energy loss:
with A a proportionality constant, p the density of the medium and kB is a factor
which Birks found to depend primarily on the type of particle. The number of scin-
tillation photons generated is proportional to the integral of Equation (A.1) over the
range of the particle's track. In the limit that < t, the scintillation light gen-
erated 2s proportional to the energy lost, and saturation only becomes important for
large dE/dx. For Compton electrons and P's, ks = (9 - 10) x 10-~ k e ~ " ' cm-2, so that dL/dx a dE/dx to within half a percent below energy losses of about 5 MeV/cm.
If the incident positron has energies greater than about 100 keV, the energy loss
will be relatively Bat at about dE/& 2 1 l\kV/cm (see Figure Al). These positrons
will only be slightly modified by the saturation effects. The energy loss of positrons
below 100 keV is larger; in this case, the Birks effect can become significant and may
affect the arnount of scintillation Light generated. As this is well below the scintillator's
threshold, we do not need to worry about these low-energy effects for the correlation
experiment . This quenching effect, however, not only occurs for the primary particle, but also
to any generated secondaries, such as b rays. The positrons in the energy range of
interest to the experiment may not be heavily affected by saturation, but it is difficult
to predict how large the dE/dx of the generated secondaries wili be, To estimate t bis
effect, GEANT simulations were performed where the Birks factor was included in the
scintillator's response. The value used was kB = 10 x 10-"ke~-~cm-* and the
results are given in Figure A 2 This is a scatter plot of the initial p's kinetic energy
against the energy lost to saturation effects, whether from the initial positron or a
FICURE Al: Total cross-sections used in GEANT for the energy loss of positrons in plastic. The dominant interaction is Bhabha (e+e- -P e+e-) scattering which generally generates 6 rays, but brernsstrahlung (e+ -. yef) and even annihiiation- in-ûight (e+e- + 27) contribute.
- O 1000 2000 3000 4000 5000
Energy deposited in scintillator [keV]
FIGURE A.2: Monte Carlo simulation of saturation effects in plastic with ks = 10 x 10'~ k e ~ " cm-2. The solid line represents the energy loss expected using Equatian (A-l), but the effect is siightly enhanced due to the production of 6 rays in the GEANT simulation.
secondary. The solid line is a calculation based on tables of the dE/dx for positrons in
NE104 plastic (only the primary positron is considered) and Equation A.1. As both
the Birks effect and 6 ray production are inherently random processes, we expect to
see the spread in the energy lost to quenching as seen in the figure. Note that for
higher /3 energies, the average energy lost is not equal to the calculation; this is due
to the higher production of d rays which, since they are lower energy electrons, suffer
more saturation than expected if no 6 rays were produced.
The Monte Carlo simulation shows that this is a relatively small effect (< 2%) and
furthemore that it is nearly linear with energy. The quenching therefore manifests
itself simply as a slight change in the slope of the energy calibration, allowing us to
neglect the efféct and assume that the light output is proportional to the energy de- posited. The spread depicted in Figure A.2 indicates that, by making this assumption
and calibrating to the average light output for a given positron energy, for any given
event we have about a f 20 keV uncertainty in the energy reading as a result of satu-
ration effects. This is not important for the correlation expeciment where we do aot
plan to utilize the event-by-event information, but analysis which incIudes kinematic
reconstructions (se, for example, jB.2) should make sure that this is considered.
Bremsstrahlung
The low-energy tail of the response functions arise mainly from three effects: (a')
particles (primary or secondary) that escape the scintiilator and therefore do not
deposit al1 of their energy, (za) radiative energy losses (bremsstrahlung) that escape
detection and (iài) annihilation-in-fiight quanta which take some of ef's kinetic energy
out of the detector. The first is brought up within chapter 4 (design simulations and
backscattering losses from the plastic) and the third is discussed in 5A.3; below we
briefly discuss bremsstrahlung losses. Often as an electron or positron is suffering collisional losses within the detector,
its velocity quickly changes when it scatters into large angles, which is enhanced due
to their small mas. This results in drastic decelerations and since any accelerat-
ing charged particle must radiate electromagnetic energy, the e* lose energy through
bremsstrahlung photons (instead of ionization losses). If these photons escape the de-
tector instead of being re-absorbed and generated into scintillation light, the detected
energy will be les than the incident particle's, thereby adding to the low-energy tail.
In the relativistic limit, which is where radiative losses contribute significantly, the
ratio of ionizing energy losses to that of bremsstrahlung for electrons is given by 1681:
For scintillators, the 2 z 5.6 so that at Te 5 5 MeV, this ratio is less than 4% (see
Figure A.3). In cornparison, this ratio is 10% for a higher Z material detector, for
example a Si(Li) for which 2 = 14, . The ratio increases because the electrons are
more likely to scatter into large angles; thus the Si(Li) bas a larger low-energy tail in its response function than the lighter scintillatort. This was one of the motivating
reasons for switching £rom a Si(Li) detector to the plastic scintiilator described in t his
thesis.
tAlso contributing ta the %(Li) low-energy tail are e* that backscatter out of the detector before stopping; this is also a fimction of the detector's 2.
-e- Be408 - f - silicon - 8- silicon k
FIGURE A.3: Radiative energy losses in plastic and silicon. Both the cross-section and average radiative energy loss is reduced in plastic compared to the higher Z silicon.
The GEANT simulations of positrons include radiative energy losses (with a cross-
section as given in Figure AJ), and also track the bremsstrahlung photon to see
if it does get reabsorbed in the plastic before escaping. To correct for differences
in electron and positron cross-sections, GEANT includes a function that scales with
T/Z2 [44]. Earlier cornparisons of measured response functions to (other) GEANT calculations [40) bad the MC consistently underestimating the low-energy tail. Al- though the response function measurements may have some systematic bias from the
e* beam characteristics, the discrepancy is a concern for us because (a) of our sensi-
tivity to the low-energy tail in the /3 - u correlation experirnent and (b) because we
have not experimentally measured the scintillator's response function and are relying
on GEANT to simulate them.
We are presently limited by backgrounds or systematics in our understanding of
the low-energy part of our 38mK spectra, and so c m o t at present use the on-line data
to test GEANT. Once these uncertainties are reduced, an attempt at measuring the
response function for 2-3 MeV positrons may be possibIe dong the slow branch (see
83.1) because the p's energy is slowly varying a t longer times-of-fiight. Reconstruction
of Ep is possible due to the overdetermined kinematics in our geometry; comparison
of the observed and calculated energies for TOF cuts dong the slow branch should at
least provide good estimates of the relative tail-to-total ratios of the response function.
Happily, preliminary calculations 1691 are in agreement with GEANT, but a dedicated
study has yet to be performed.
Annihilation Radiation
Up to now, the effects have been applicable to both electrons and positrons, with
perhaps only minor differences in the details. The response function of positrons will
have additionai components unique to them because they annihilate with free and
atomic electrons.
Compton summing In the case that the positron cornes to rest before annihilating
with a free electron, we know from simple energy-momentum conservation that two back-to-back 7 s will be generated with momenta p,, = -p , and ET, = E, =.me. If these annihilation photons do not interact with the scintillator, then the response is unaffected since the positron deposited al1 of its kinetic energy. However, if one
Compton scatters within the scintillator, then the additional energy deposited by
the Compton scattered electron will be added to the positron's signal. The positron
response will therefore have a high-energy tail extending up to 340 keV from the
full-eaergy peak. Although les likely, both annihilation 7s rnay Compton scatter
cxtending this taii even further, up to 680 keV. This 'Compton toe' is a concern
in the correlation experiment because an improper calculation of this high-energy
Compton tail d l afiect the scintiliator's calibration; the 5.022 MeV end point i d1
appear shifted if too much or not enough Compton sumrning is included.
Annihilation-in-Bight An additionai complication with positrons is the fact that
they can annihilate before coming to rest (annihilation-in-8ight). The ys d l no
longer be back-to-back and E, # E, #me as some of the kinetic energy of the
positron is transferred to annihilation radiation. This t h e , if the annihilation ys
escape the detector, the energy reading will be in the low-energy tail of the response
function because the positron did not deposit al1 of its energy in the detector before annihilating. The cross-section for annihilation into two 7s is given by [70]f
with T, = 2.818 Çi and 7 = E/m,. This cross-section is compared with that of
collisional losses in Figure A.1 and, though it is small compared to ionization losses in our energy range, the integrated probability of annihilation-in-flight is as large as
10% for a 5 MeV positron [711.
t.i\nniLilation to one 7 can occur if the electron is bound, but this cross section goes like (aQ3 compared to a* and so only contributes in hi& Z materiais.
GEANT and Future Work
The success of GEANT in reproducing the observed ,O spectra enables us to place some
trust in its simulations. This coupled with its ease of prograrnming to suit a users
needs has meant it has found applications to other aspects of TRINAT'S potassium
program. Two of these applications, improved geometry designs to reduce scattering
and searching for massive neutrinos, are described below.
B.1 Future geometries
In order to obtain a more uniform and calculable electric field, a redesign of the
electrostatic hoop system is being performed and is hoped to be implemented before
the next running period. Inspired in large part by the scattering effects presented
earlier (see Section 4.5.1), design simulations of the experiment are concurrently being
performed to try to minimize these effects in the new geometry.
One of the major changes to the electric hoops (see Figure B.1) is in the one closest
to the P-telescope. In order to be less sensitive to where ground is defined, this hoop
has been changed into two wide, Bat concentric rings ('hoop 5' and the collimator)
maintaineci at different potentials. On the other side, another end plate ('hoop O*)
has also been added for more uniform fields on the recoil detector's side of the trap.
Then the natural questions became ' m a t sizes (inner diameters and thicknesses) and
materials (Be, Al, W, Ta) should we choose for these new volumes?" The GEANT geometry was amended to include hoop 5 and collimatorso that simulations could help
us to best answer these questions. The geometry was further adapted by T.J. Stocki,
a research associate working with TRINAT, to reflect the rest of the changes in the
FIGURE B.l: Schematic diagram of TRINAT'S new electrostatic hoop design. The changes can be seen by comparing this to Figure 4.7 on page 40.
new electrostatic hoop design, including hoop O and the ceramic rods used to mount
the hoops.
Simulations were run for various dimensions and materials of the collimator and
hoop 5; the MC indicates a factor of two improvement over the geometry presented in
this thesis if a copper-tungsten collimator is used to restrict the cone of ,h from the
trap to hit only the foi1 of the /3 window (and not the steel of the front face). Hoop 5
is made of glassy carbon, and it was found that a number of ps scattered through
it and then into the P-telescope; in an effort to reduce these new scattering effects,
another copper-tungsten collimator (the ' P window collar') was added to the front
face of the scintillator vacuum chamber. The collar extends out far enough that ps
have no solid angle for entering the telescope directly from hoop 5 and greatly reduces
these scattered events.
The material of hoop O was carefully considered because we want to make sure
we minimize backscattering effects. If this plate is also made of glassy carbon, the 0s
have a good probability of transmitting through the plate rather than backscattering;
once through, they readily scatter off the other (higher 2) materials, but the solid
angle for firing the P-telescope is very small.
B.2 MASSIVE NEUTRINOS 119
B.2 Massive neutrinos
The correlation experiment, as mentioned earlier, provides us with the momenta (both
direction and magnitude) of both the recoil and the B from the 3BmK decay. From
these measurements, we can deduce the neutrino momentum on an event-by-event
basis, limited by how well the ef energy is measured. The kinematics of the decay
described in 53.1 (on page 14) assumed a massless neutrino, and indeed we know this
to be tme at the eV level for the v,. The limits on the C( and T neutrino's, however,
are not as stringent and, if they have mass, the observable eigenstates of the weak
interaction can mix with the mass eigenstates. In this case the kinematics of the decay
will be quite different, as depicted in Figure B.2. For the fast branch, where there
are many events, a massive neutrino will take energy away from the recoil to conserve
momentum, resulting in an extra ridge at longer time-of-flight . The separation of this
ridge from the fast branch where no mixing occured will be related to the mass of the heavy neutrino; for a 2 MeV heavy neutrino in the geometry of this experiment, it
corresponds to a shift of approximately 200 ns. The relative population of this ridge compared to that of the regular fast branch is determined by the probability of the
electron neutrino mking with a heavier one (as well as changes in the phase space of
the decay and the angular correlat ion).
With the 38mK decay Q-value of 5.022 MeV, our geometry should be sensitive to
neutrinos in mass range of approximately 1-4 MeV. The analysis, which is the central
part of M. Trinczek's Ph.D. thesis [721, will reconstruct the recoil TOF, assuming a
rnassless neutrino, as a function of Tp bins, and then compare tbis cakulated time with
the observed TOF. The massive neutrinos will be clearly visible as their recoostructed
TOF will be incorrect; for this reason, chey will be far removed in TOF from the fast
branch's peak at TOFOb - TOFdc = O. The overall number of counts in each of the
h o peaks wiii be used to place limits on the mixing strength of v, --+ vh,,,.
In order to be able to accurately fit the (regular) fast branch's peak, a MC simula-
tion must be performed to properly account for (a) various scattering effects, ( i i ) the
h i t e trap and detector sizes, and ( i i à ) the response function of the 0-telescope. MC simulations 1651 where the recoil and positron are tracked with a simple mode1 of ( i )
(at most one scatter) are much faster than GEANT simulations (= 4000~). This fast
B.2 MASSIVE NEUTRINOS 120
- - mv = 1 MeV -- . mw = 3 MeV mv = 2 MeV - m v = 4 M e V
FIGURE 8.2: Kinernatics of 3 8 m ~ decay where the emitted neutrino is massive (dashed lines), for masses m, = 1,2,3 and 4 h4eV; the solid line is m, = O. The position of the heavy neutrino ridges depends on its mass, and the relative intensity of this ndge compared to that of the regular fast branch is related to the mixing strength.
simulation accounts for ( i i ) and includes (iii) by convoluting the generated 0 energy
with GEANT simulations of the telescope's response functiont.
The detailed tracking of the positron using GEANT, however, appears to be es-
sentiai because the fast simulations fail to reproduce the tail of the main TOFOb, - TOFCd, = O peak [731. One major source of the TOF tail is the (back)scattered ps because the reconstnicted TOF d l have a wrong d u e of Op, in the calculations.
The massive neutrino peak, though shifted in TOF from the main peak, will still be
in the tail of the m, = O peak. Analysis of the reconstructed events, which looks
for very small peaks on top of the large zero m a s 'background,' therefore requires
GEANT simulations since they best reproduce the m, = O tails.
The m, > O spectrum shape does not need to be as accurately for the andysis
because it is so small compared to the main m, = O peak. For this reason, the fast
Monte Carlo is being used to generate the massive neutrino events and GEANT is only
tThese tespanse functions m e generated in 1998, and so updating them should help to irnprove the fast MC
B.2 MASSIVE NEUTRINOS 121
be needed to generate the background.
Electronics of the fi-Telescope
A schematic of the electronics diagram for the telescope assembly is depicted in Fig-
ure C.1. Below is a brief overview of the system as a whole.
The DSSSD provides us with 24 Z + 24 jj = 48 energy signals and 6 groups of $
timing signals (consisting of 4 strips each). Each of the 24 y-strip's timing signals have discriminators set to 5 15 keV; if any of the strips p a s this hardware threshold,
the timing signai is used to generate a (wide) DSSSD event trigger. The gain of al1 48 strips is monitored using the Ortec 448 research pulser.
The scintillator energy is taken £rom the dynode output of the PMT after it is
invected by the LeCroy 428F linear fan-in/fan-out. The high voltage applied to the
PMT is adjusted by the stabilization unit based on the LED signals observed; the
intensity of the LED is (independently) maintained constant using a temperature stabilized photodiode. The Tennelec 455 constant fraction discriminator uses the
anode signai to generate the scintiljator's (short) timing signal. The gates for the ADCs and the starts for the TDCs are al1 derived frorn the
LeCroy 429A unit labeiied 'event trigger'. The inputs to this unit, i.e. the various event types, are:
1. A prescaled scintiiiator event,
2. A hardware DSSSD-scintiilator coincidence,
3. DSSSD/MCP pulser events Gom the Ortec 448,
4. Scintillator pulser (LED) events fiom the stabilization unit, and
APPENDM C ELECTRONICS OF THE FTELESCOPE 123
FIGURE (2.1: Electronics diagram for the DSSSD and scintillator.
APPENDIX C ELECTRONICS OF THE ~TELESCOPE 124
5. Event triggers from the MCP (though not generally used on-line because of its
high rates).
The coincidence register (C212) allows us to record which of these generated the
event tcïgger. In addition, this unit is used to define where within the trap cycle a
given eveat took place; the signais are taken from the PC controlling the trap and
tell us when the atoms are being transferred (push beam on/off), if there are atoms
in the trap or if they are being loaded (trap on/off) and what detunings are used in the MOT'S lasersf ('tiny' trap on/ofT).
The CAMAC system is used to acquire the data and is recorded in the YBOS format both on the disk of the host Pentium III 500 MHz Linux PC as well as on
magnetic tapes. Both the on-line and off-line data malysis are done using a TR~uMF standard program [741 with a special subroutine added to incorporate the analysis
scherne of the DSSSD 4.2.5.
tBy changing the detunings once the trap is already loaded, the size of the atom cloud is ceduced.
Bibliography
(11 F. Halzen and A.D. Martin. Quarks and Leptons: An Introductory Course in Particle Physics. John Wiley and Sons, 1984.
121 K.S. Krane. Introductory Nuclear Physics. John Wiley and Sons, 1988.
131 D. Grifiiths. Introduction to Elementary Particles, chapter 10.7. John Wiley and
Sons, 1987.
[4I E. Fermi. Zeitschrift für Physik, 88, (1934) 161.
[51 The Particle Data Group. Review of particle physics. The European Physical
Journal, C3, (1998).
[6) M A . Preston. Phystcs of the Nucleus, chapter 15, page 424. Addison-Wesley,
1965. He cites (41 as a reference.
[71 T.D. Lee and C.N. Yang. Question of parity conservation in weak interactions.
Physical Review, 104, (1956) 254-258.
[SI The Hl Coliaboration. A search for leptoquarks at BERA. Physics Letters, B369,
(1996) 173-185.
[9I The DO Collaboration. Search for first genemtion scalar leptoquark pairs in collisions ut fi = 1.8 TeV. Physical Review Letters, 80, (1998) 2051-2056.
[lOI E.G. Adelberger, C. Ortiz, A, Garcia, H.E. Swanson, M. Beck, O. Tengblacl,
M.J.G. Borge, 1. Martel, H. Bichsel, and the ISOLDE Collaboration. Positron-
neutrino correlat2ons in the Of-, O+ decay of 32Ar. Physical Review Letters,
83, (1999) 1299-1302.
(111 A.S. Carnoy, J. Deutsch, and P. Quin. The supemllowed Of+ O+ beta-
transitions and possible scalar wupling. Nuclear Physics, A568, (1994) 265-270.
[121 J.D. Jackson, S.B. Trieman, and H.W. Wyld. Possible tests of time reversal
invariance in beta decay. Physical Review, 106, (1957) 517-521.
1131 J.D. Jackson, S.B. Trieman, and H.W. Wyld. Coulomb corrections in allowed
beta transitions. Nuclear Physics, A4, (1957) 206-212.
[141 P.M. Endt. Supplement to energy levels oJ A = 21 - 44 nuclei (VII). Nuclear Physics, A633, (1998) 1-220. 37K on page 180; 38*38mK on page 187.
[151 E. Hagberg, V.T. Koslowsky, J.C. Hardy, LS. Towner, J.G. Hykawy, G. Savard, and T. Shinozuka. Test 01 isospàn rnixing corrections in superallowed
O+* Of decays. Physical Review Letters, 73, (1994) 396-399. (See also Er- ratum: Physical Review Letters, 74, p. 1041).
[161 P.D. Harty, N.S. Bowden, P.H. Barker, and P.A. Amundsen. Energy of the
superal/owed P decay of 38mK. Physical Review, C58, (1998) 821-25.
[171 H. Trottier, 2000. Private communication.
[181 A. Gorelov. PhD thesis, Simon Fraser University. In preparation.
119) E.L. Raab, M. Prentiss, A. Cable, S. Chu, and D.E. Pritchard. Tmpping of neutral sodium atoms with radiation pressure. Physical Review Letters, 59, (1987) 2631-
2634.
[201 H. Metcalf and P. van der Straten. Coolàng and tmpping ojneutral atoms. Physics Reports, 244, (1994) 203-286.
[211 G.D. Sprouse and L.A. Orozco. Laser trcrppàng of radioactive atoms. .!4nnual Reviews of Nuclear and Particle Science, 47, (1997) 429-461.
[221 E. Arimondo, W.D. Phillips, and F. Stnimia, editors. Laser Manipulation of
Atoms and Ions. Amsterdam: North Holland, 1992.
[231 D.R. Swenson and L.W. Anderson. Relaxation mies for optically pumped Na vapor on silicone surfaces. Nuclear Instrumentation and Methods, B29, (1988)
627-642.
[241 G. Ball, D.M. Moltz, J. Cerny, J. Powell, G. Savard, J.C. Hardy, S. Bishop, J.M. D'Auria, P. Bricault, J.A. Macdonald, J.R. Leslie, H.-B. Mak, and I.A. Towner.
Pure Fermi decay in medium m a s nuclei, 2000. TRIUIJIF experiment E823.
1251 K.S. Krane, B.G. Turrell, J.L. Wood, J.M. D'Auria, B. Heinrich, C.A. Davis, P.P.J. Delheij, P. Mantica, R. Kiefl, A. Kotlicki, J. Pond, R.C.C. Ward, and
M.R. Wells. Nuclear moments in the mass-100 region, 2000. TRIULIF experiment E828.
[261 S. Dunsiger, R. R/liller, R.F. Kiefl, J-C. Chakhalian, W.A. MacFarlane, J. Pond, J. Sonier, B. Turrell, B.D. Gaulin, M. Gingras, R. Heffner, B. Heinrich, and
B. Ittermann. 0-nmr investigations of magnetic multilayers and giant magne- toresistance, 2000. TRIUWF experiment E815.
[271 R.N. Boyd, L. Buchmann, JM. D'Auria, J.D. King, and 1. Tanihata. Proposed study of the 'Li(a, n)"Be reaction, 2000. TRIUMF experiment E812.
(281 N.P.T. Bateman, J.M. D'Aucïa, and the DRAGON collaboration. Measurement of the astrophysical rate of the 21N~(p, y)22Mg reaction, 2000. TRIUMF experiment
E824.
[291 J.M. D'Auria, L. Buchmann, M. Dombsky, P. MCNeely, G. Roy, H. Sprenger, and
J. Vincent. Upgrade of the TRIULIF on-fine asotope separutor, TISOL. Nuclear Instrumentation and Met hods, 870, (1992) 75-79.
[301 J.M. D'-4uria, J. Behr, L. Buchmann, M. Dombsky, K.P. Jackson, and H. Sprenger. The TISOL facility at TRIUMF: operational statvs ut 10 years. Nuclear Instrumentation and Methods, B126, (1997) 7-11.
[31] ISOLDE user's guide. Technical Report 8605, CERN, 1986.
[321 J.A. Behr, A. Gorelov, T. Swanson, O. H5usser, K.P. Jackson, hl. Trinczek, U. Giesen, J.M. D'Auria, R. Hardy, T. Wilson, P Choboter, F, Leblond, L. Buch- m a n , M. Dornbsky, C.D.P. Levy, G. Roy, B.A. Brown, and J. Dilling. Magneto-
optic tmpping of /3-decaying S8mK, 37K h m an on-line isotope separator. Physicai Review Letters, 79, (1997) 375-378.
1331 J. Dilling, J.A. Behr, A. Gorelov, T. Swanson, O. Hausser, D. Melconian, K.P. Jackson, M. Trinczek, U. Giesen, and J.M. D'Auria. 0-decuy eqeriments 01 neutml atorns 3n a magneto-optic trap. Hyperiine Interactions, 115, (1998) 159-
164.
1341 J.A. Behr, A. Gorebv, D. Melconian, M. Trinczek, P. Dubé, O. Hausser, U. Giesen, K.P. Jackson, T. Swanson, J.M. D'Auria, M. Dombsky, G. Baii, L. Buchmann, B. Jennings, J. Dilling, J. Schmid, D. Ashery, J. Deutsch, W.P. Al- ford, D. Asgeirsson, W, Wong, and B. Lee. Beta-neutrino correlation ezperiments
on laser trapped 38mK, 37K. Hyperüne interactions. (in press).
(351 P. Bricault, M. Dornbsky, P.W. Schmor, and G. Stanford. Radioactive ion beams
facility at TRIUMF. Nuclear Instrumentaion and Methods, B126, (1997) 231- 235.
[361 R. Gückert, X. Zhao, S.G. Crane, A. Hime, W.A. Taylor, D. Tupa, D.J. Vieira, and H. Wollnik. Magneto-optical trapping O/ nidioactiue *ZRb atoms. Physical Review (rapid communications), A58, (1998) R1637-R1640.
[37] T. Swanson, D. Asgeirrson, J.A. Behr, A. Gorelov, and D. Melconian. Eficient
tmnsj?r in a double rnagneto-optic irap system. Journal of the Optical Society of America, B 15, (1998) 2641-2645.
[38j Galileo. Galileo Micro-Channel Plate data sheets, The recoii detector is largely the work of A. Gorelov.
[391 B. Brehm, J. Grosser, T. Ruscheinski, and M. Zimmer. Absolute detectlon efi-
ciencies of a microchanne1 plate detector for ions. Measurements in Science and
Technology, 6, (1995) 953-958. See also J. Oberheide, P. Wiihelms and M. Zirn- mer, New results on the absolute ion detection eflcaencies of a microchannel
plate. Measurements in Science and Technology, 8, (1997) 351-354.
1401 J. Dilling. Beta-zerfall untersuchungen von 37K und 38"K in einer neutralen atom
laser falle. Diplomarbeit thesis, Ruprecht-CaclsUniversitat Heidelberg, 1997.
[dl] O. Hausser. B asymmetry rneasurements using 37K. (unpublished), 1998.
[421 E.T. Clifford. Kinematic sshifts in @ detayed prticle ernission as a probe of0 - u angular comlatiom, PhD thesis, University of Toronto, 1981.
1431 E.T. Clifford, E. Hagberg, V.T. Koslowsky, J.C. Hardy, H. Schmeing, and R-E. Azuma. Meusuremelits of the response /unction of a hybrid debedor telescope to
monoenergetic beams ojpoQitroa~ and electrons in the energy range 0.8-3.8 MeV. Nuclear Instrumentation and Methods, 224, (1984) 440-447.
[44j CERN. GEANT Detector Description and Simulation Tool, October 1994 edition,
1993.
[451 A. Gorelov and D. Melconian, 1999. The careful and meticulous design of the
DSSSD mount was done by Alexandre.
1461 W.T. Scott. The theory ofsmall-angle niult2ple scattering of /ast charged particles.
Reviews of Modern Physics, 35, (1963) 231-313.
(471 Brush Wellman Inc. Muterial specOJication sheet. The type of beryllium decided
upon was El.
[48j Bicron Corp. Bicron data slieet, 1997.
[491 U. Giesen. TRINAT DSSSD electronics. Locally designeci and constructed, but
largely based upon the Hemes electronics., 1994.
(501 C.R. Huclbut. Plastic Scintillators: A Survey. Bicron Corp., 1985. Presented at
the -i\merican Nucleai- Society W i t e r Meeting, November, 1985.
[511 T. Davinson, A.C. Shotter, E.W. MacDonald, S.V. Springham, D. Jobanputra,
A.J. Stephens, and S.L. Thomas. Deuelopment of a silicon strip detector a m y jor nuclear stmcture phpics. Nuclear Instrumentation and Methods, A288, (1990)
245-249.
[52) Micron Semiconductor Ltd. Micron Semiconductor data sheet, 1998. BB2-500 design.
[53j R.A.B. Shaheen. Nuclear data sheets for A = 133. Nuclear Data Sheet, 75,
(1995) 491.
1541 Y.A. Ellis-Akovali. Nuclear data sheets for A = 241. Nuclear Data Sheet, 74,
(1995) 461.
1551 J. Yorkston, A.C. Shotter, D.B. Syme, and G. Huxtable. Interstrip surface eflects
in ozide passiuated ion-implanted silicon strip detectors. Nuclear Instrumentation
and Methods, A262, (1987) 353-358.
1561 P.V. Vavilov. Ionization Iosses of high energy heauy particles. Soviet Physics
JETP, 5, (1957) 749.
1571 A. Gorelov, 2000. Private communication,
[58) M.J. Martin. Nuclear data sheets for A = 207. Nuclear Data Sheet, 70, (1993) 315.
[591 C. Moisan. Private communication. Diffuse reflectors were Found to yield the
best light collection for the PET scintillators.
1601 O. Klein and Y. Nishina. 2. Phys., 52, (1929) 853.
(611 Philips Photonics. Photomultiplier Tubes: Principles 6 Applacations, 1994. see
Equation (3.34).
[62j Y. Hoiler, J. Koch, and A. Naini. A stabilized NE213 scintillator !or neutron
time-of-Pight spectroscopy. NucIear Instrumentation and Methods, 204, (1983) 485-490-
163) A. Gorelov, 199û-present . The decay generation and recoii tracking subroutine, beta,decay.F, is taken from the 'fast Monte Carlo' [651.
[64] E.T. Clifford, 1999. Privat e communication.
[65) A. Gorelov, 2000. Monte Carlo simulation program (Fortran) for TRINAT'S P - v correlation experiment.
1661 W. Bauer, J. Briggman, H.-D, Carstanjen, S. Connell, W. Decker, J. Diehl,
K. Maier, H.-E. Schaefer, A. Seeger, H. Stoil, and E. Widmann. The stuttgart positron beam, its performance and m e n t experiments. Nuclear Instrumentation
and Methods, B50, (1990) 300-306.
1671 J.B. Birks. Theoy and practice O/ scintillation counting. Pergamon Press, 1964.
1681 R.D. Evans. The Atomic Nucleus. Robert E. Kreiger Publishing Company, 1955.
[691 J.A. Behr, 1999. Private communication.
(701 W. Heitler. The Quantum Theoy of Radiata'on. Clarendon Press, Oxford, 1954.
1711 R.B. Firestone and V.S. Shirley, editors. Table of Isotopes, volume I I . John Wiley and Sons, 8th edition, 1996.
[721 M. Trinczek. PhD thesis, Simon Raser University. In preparation.
[731 M. Trinczek, 2000. Private communication.
1741 P.W. Green. NOVA. Trubw / The University of Edmonton, v2.0 edition, 1995.